Edge Computing - From Hype To Reality PDF
Edge Computing - From Hype To Reality PDF
Edge Computing - From Hype To Reality PDF
Edge
Computing
From Hype to Reality
EAI/Springer Innovations in Communication
and Computing
Series editor
Imrich Chlamtac, CreateNet, Trento, Italy
Editor’s Note
The impact of information technologies is creating a new world yet not fully
understood. The extent and speed of economic, life style and social changes
already perceived in everyday life is hard to estimate without understanding the
technological driving forces behind it. This series presents contributed volumes
featuring the latest research and development in the various information engineering
technologies that play a key role in this process.
The range of topics, focusing primarily on communications and computing engi-
neering include, but are not limited to, wireless networks; mobile communication;
design and learning; gaming; interaction; e-health and pervasive healthcare; energy
management; smart grids; internet of things; cognitive radio networks; computation;
cloud computing; ubiquitous connectivity, and in mode general smart living, smart
cities, Internet of Things and more. The series publishes a combination of expanded
papers selected from hosted and sponsored European Alliance for Innovation (EAI)
conferences that present cutting edge, global research as well as provide new
perspectives on traditional related engineering fields. This content, complemented
with open calls for contribution of book titles and individual chapters, together
maintain Springer’s and EAI’s high standards of academic excellence. The audi-
ence for the books consists of researchers, industry professionals, advanced level
students as well as practitioners in related fields of activity include information and
communication specialists, security experts, economists, urban planners, doctors,
and in general representatives in all those walks of life affected ad contributing to
the information revolution.
About EAI
Edge Computing
From Hype to Reality
123
Editor
Fadi Al-Turjman
Computer Engineering Department
Antalya Bilim University
Antalya, Turkey
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To my great parents, my first lady and my
little one.
Whose affection, pray, and continues
encouragement have made this book and
every beautiful thing in my life realized.
Fadi Al-Turjman
Preface
We are living in an era where the Edge of the Cloud is becoming a global platform
for the computation and interaction between machines and smart objects, in real-
time applications and for many critical tasks.
With the application areas such as smart cities, smart grids, smart cars, smart
eHealth, smart supply chain, and smart homes in the Internet of Things (IoT), we can
consider the Edge computing as a complete package of the smart networked objects.
From this perspective, it is essential to understand the role of Edge computing which
will provide a global backbone for the worldwide information sharing/processing in
the near future.
There is no doubt that introducing such a new phenomenon can come with
potential challenges especially in terms of security and quality of service (QoS).
Therefore, it is essential to consider new enabling technologies such as wireless
sensor networks (WSN), various radio technologies and cellular infrastructures
(e.g., small cells), radio-frequency identification (RFID), and cloud services and
architectures for performance optimization.
The objective of this book is to present a survey of existing techniques and
architectures for Edge computing optimization and improvements with the presence
of IoT systems. The main focus is on practical design aspects that can help in
realizing such a paradigm in practice. The applications of Edge computing in
IoT, evaluation metrics, constraints, and open issues about the addressed topic are
included for discussion as well. This conceptual book, which is unique in the field,
will assist researchers and professionals working in the area of the Cloud to better
assess the proposed Edge paradigms that are already beginning to be a significant
part of the global infrastructure.
vii
Contents
1 The Era of the Personal Cloud: What Does It Mean for Cloud
Providers? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
Mohamed Abu Sharkh, Abdallah Shami, and Mohamad Kalil
2 Optimization in Edge Computing and Small-Cell Networks . . . . . . . . . . 17
Jitender Grover and Ram Murthy Garimella
3 A Comprehensive Survey on Architecture for Big Data
Processing in Mobile Edge Computing Environments . . . . . . . . . . . . . . . . . 33
Maninder Jeet Kaur
4 Taxonomy of Edge Computing: Challenges, Opportunities,
and Data Reduction Methods . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 51
Kusumlata Jain and Smaranika Mohapatra
5 Applications of the Internet of Things with the Cloud
Computing Technologies: A Review .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
U. Ram Jagannath, S. Saravanan, and S. Kanimozhi Suguna
6 Software-Defined Internet of Things to Analyze Big Data
in Smart Cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 91
Sadia Din, Awais Ahmad, Anand Paul, and Gwanggil Jeon
7 Taking Cloud Computing to the Extreme Edge: A Review
of Mist Computing for Smart Cities and Industry 4.0 in Africa . . . . . . 107
Eustace M. Dogo, Abdulazeez Femi Salami, Clinton O. Aigbavboa,
and Thembinkosi Nkonyana
8 IoT and Edge Computing as a Tool for Bowel
Activity Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 133
Umit Deniz Ulusar, Erdinc Turk, Ahmet Sefa Oztas,
Alp Erkan Savli, Guner Ogunc, and Murat Canpolat
ix
x Contents
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 185
Chapter 1
The Era of the Personal Cloud:
What Does It Mean for Cloud Providers?
1.1 Introduction
covers a transfer away from the full range of personal computing usage patterns
to a different set of patterns using Cloud services. For example, Gartner estimated
that consumers’ storage will grow from around 500 gigabytes (2011 estimate) to 3.3
terabytes by 2016. A third of these loads is already in the Cloud [22].
As for network loads, 2014 was the first year a majority of the world data
workloads were on the Cloud (51% to 49% for data centers). Cisco expected the
growth of this number to be 63% in 2017. Moreover, the estimate for network
traffic was 69% for Cloud networking compared with 31% with regular data centers.
Mobile date traffic is a slice that is affected largely as well [16]. Cloud mobile data
traffic is projected to grow from 81% in 2014 to 90% in 2019.
Research firm IDC estimates that nearly one third of the worldwide enterprise
application market will be SaaS-based by 2018, driving annual SaaS revenue to
$50.8B, from 22.6B in 2013 [17]. According to a survey reported in [34], the average
enterprise uses 1,427 distinct Cloud services with the average employee actively
using 36 Cloud services at work. Gartner estimated the worldwide public Cloud
services market at $209.2B and predicted this will pass the $250B barrier soon [23].
The Internet of Things offerings expand the horizons of data sources for a certain
Cloud-based service, but it also jumps to a new dimension in terms of the data
filtering and preprocessing required of the same service to get to the core value
returning requests. The sheer volume of the connected devices along with the
connection speed and user expectations for response from the server put a stress
both on the network and on back end Cloud data center to perform up to the required
level. This poses a challenge to Cloud service designers regarding how to optimize
this process. You cannot expect every single data packet from 8.4B heterogeneous
IoT devices (Gartner’s estimate of the number of devices deployed around the world
in 2017) to be informative or even relevant.
Personal Cloud spread contributed to the Cloud market size multiplication. The
consumer’s interest combined with technological advancements contributed to these
leaps in consumption. Public Clouds – which are where most of the personal Cloud
loads reside – are the most affected with this growth. Public Cloud vendor revenue
grew from 26 billion dollars in 2012 to a projected 154 billion in 2017 [45].
Software as a service represented 67% of this as of 2015. Traditional IT
infrastructure is definitely losing ground to Cloud solutions. The market share
percentages moved swiftly from 72%, 17%, and 11% for traditional IT, public
Clouds, and private Clouds, respectively, in 2011 to 53%, 29%, and 5% in 2017 [43].
Personal work loads are going in a clear direction: up to the Clouds.
In this work, we endeavor to investigate the potential impact the dawn of this new
era will have on Cloud providers. The purpose is to offer a solid perspective into
what changes are expected in the environment, technical requirements and usage
trends. This, in turn, would aid in forming a vision for the directions providers need
to go into their policies and algorithm planning.
In Sect. 1.2, we investigate the impacts on the area of resource allocation in a
Cloud data center. This covers multiple venues including the request distribution
effects, impacts on financial investment in Cloud resources, the optimality of the
pricing plans, and energy efficiency considerations in light of the personal Cloud
1 The Era of the Personal Cloud: What Does It Mean for Cloud Providers? 3
conditions. Section 1.3 offers a detailed discussion on Cloud client demand pattern
prediction methods.
In Sect. 1.4, we introduce the impacts and the intersection points with another
dominating technology trend, which is the Internet of Things. This section covers
the challenges expected from numerous nodes representing the clients in that
scenario. The cause and effect of the diversity and volume of these requests is probed
into.
Section 1.5 strives to envision any tangible effects on Cloud brokering services.
Finally, venues for future improvement are explored in Sect. 1.6.
For a Cloud provider, the resource allocation is in the core of the Cloud management
process. Resource utilization and long-term planning can be really lucrative in
terms of performance and financial rewards. This is generally easier to handle with
enterprise clients. Despite Cloud computing having the major advantage of dynamic
scaling and pay as you go as a benefit for clients, it is noticeable that precise
prediction of the resource scale the client will demand would be rewarding for the
client. For a large enterprise, the favorable scenario would be to rent an amount
of Cloud resources as close as possible to the amount they practically require on
a long-term contract with the ability to scale up or out just in case demand surges
happen. As depicted in detail later, dynamic scaling is a business-enhancing feature,
but it being the norm does not make sense cost-wise.
A look at some of the public Cloud price offerings confirms this idea. It is a
practical and realistic assumption to expect enterprise clients to have the expected
average load and maximum load estimates calculated and clear. However, with
clients using personal Clouds or services with higher exposure to individuals using
their own devices, this can get much trickier. For example, if we look at pricing
patterns for VM instance rental for Amazon AWS and if we focus on infrastructure
as a service since it often is the clearest example of the resource outsourcing and
economies of scale, some interesting insights can be drawn [7].
Amazon’s three charging models are on-demand, which, as the name suggests,
basically reflects a pay-as-you-go system. Understandably, instances under this
usage condition tend to be of higher price and priced per hour. Next is the reserved
instances which provide the clients with guaranteed capacity for a fixed term. These
are offered with (no upfront, partial upfront, and all upfront) payment options and
on a 1-year or 3-year terms. Finally, Amazon offers the clients the opportunity to bid
for the remaining unrented capacities in a specific Clouddata center. To buy these
instances – called spot instances – clients determine their demands and put a bid in
4 M. Abu Sharkh et al.
0.3000
3Yr-All-Upfront-Resevred
1Yr-All-Upfront-Resevred
0.2500
On Demand
Spot
0.2000
0.1500
0.1000
0.0500
0.0000
Fig. 1.1 A comparison of Cloud service prices per hour for a select set of instances offered by
Amazon under different charging modes [7]
the form of the price they are willing to pay, and then the instances are sold subject
to the available capacity in a data center.
Figure 1.1 shows a comparison of the prices of 4 pricing options (3-year reserved,
1 year reserved, on-demand, and spot instances based on the latest accepted bids) as
of Feb 2016 for several instance types. We chose to use Linux-based instances for
this comparison (since the environment change affects the price per hour). Each
instance type reflects a different resource configuration. This figure along with
Fig. 1.2 – which shows the percentage of the on-demand instance price for the other
three pricing methods – illustrates the significance of the usage condition to the
client. As for the provider side, a look at the revenue that can be generated from a
specific load for a Cloud with predetermined capacity can really help us achieve and
form a trend here.
The variables in play here are the prices per hour for each usage condition. α
is defined to be the price per hour for 3-year reserved instances. β represents the
price per hour for 1-year reserved instances. γ represents the price per hour for
the on-demand instances. δ represents the price per hour for the instances sold in
the spot mode based on the latest spot instance price. X represents the number of
3-year reserved instances. Y represents the number of 1-year reserved instances.
Z represents the number of on-demand instances. Finally, W is the number of
instances sold in spot mode in a specific hour. The hourly revenue generated from a
specific can be calculated as follows.
REVh = α × Xh + β × Yh − γ × Zh + δ × Wh (1.1)
In Fig. 1.3, the expected revenue for a Cloud provider from a specific data center
is shown. The data center used is assumed to have a capacity of 1000 VM instances.
1 The Era of the Personal Cloud: What Does It Mean for Cloud Providers? 5
70
60
50
40
3Yr-All-Upfront-Resevred
30
1Yr-All-Upfront-Resevred
20
Spot
10
Fig. 1.2 A comparison of Cloud service prices per hour relative to the on-demand pay-as-you-go
charging model for a select set of instances offered by Amazon [7]
14.0000
12.0000
10.0000
3Yr-All-Upfront-
8.0000 Resevred
1Yr-All-Upfront-
6.0000
Resevred
4.0000 on demand
2.0000
0.0000
Fig. 1.3 The total revenue achieved by a Cloud of 1000 VM instance capacity under different
client demands
The figure represents the expected revenue for a Cloud provider when different
types of clients rent instances of the t2.micro type. For each client, revenue is
calculated when the demand ranges from 1000 (full Cloud capacity) until 0. In
the case of the reserved usage conditions (3 year and 1 year), the original contract
is assumed to be 800 instances with any increase rented by the client on an on-
demand condition. This figure tells us that with the scattered on-demand pattern
which best describes the personal Cloud requests, Cloud providers end up having to
auction the remaining capacity as spot instances. This could make it harder for the
provider to reach the revenue level achieved when clients use reserved mode. In the
6 M. Abu Sharkh et al.
figure, the utilization level where the direction/trend changes is (according to this
scenario of posted prices) the point where 45% of the data center requests being on-
demand. Otherwise, it is evidently better for the provider to reserve the data center
resource for the less expensive yet more stable reserved usage condition clients.
These are mostly enterprises. This all add extra emphasis on the Cloud providers
traffic prediction and resource reservation policies in the case of heavy personal
demand. More focus on this area would aid the providers in mitigating the effect
of other factors impacting the pricing/commercial side (competitors, client global
demand, saturation levels, government-related factors in terms of regulation and
taxation).
Power consumption in Cloud data centers is a pressing issue for Cloud providers.
Power costs represent between 25% and 40% of the operational expenses of a
data center [47]. The Natural Resources Defense Council (NDRC) published a
data center efficiency assessment in Aug 2014 as an attempt to depict the scale
of data centers the world over [36]. The study mentions that US data centers are
on track to consume roughly 140 billion kilowatt-hours of electricity annually by
2020, equivalent to the output of 50 large power plants (each with 500 megawatts
capacity). If worldwide data centers were a country, they would be the globe’s
12th largest consumer of electricity. Another fact here is their assessment of
energy efficiency. Their analysis finds that up to 30% of servers are obsolete or
not needed and that other machines are grossly underutilized. Persistent issues
obscuring efficiency include: peak provisioning, failure to power down unused
servers, and competing priorities, keeping costs low and maintaining high levels
of security, reliability, and uptime for their clients. These are some of the main
factors cited that affect power efficiency and stand in the way of a staggering 40%
potential improvement in power consumption. These factors are all largely affected
by data center load planning and management. An efficient scheduling energy-aware
algorithm would exploit the benefits that come from virtualization technologies,
optimal demand-driven provisioning, and efficient load modeling.
In the era of personal Cloud, different energy efficiency algorithms will have
added level of complexity to tackle [2, 13, 15]. As discussed in detail in [1], energy
efficiency algorithms take one of three routes: server consolidation [10, 24, 42]
and switching off of idle server, dynamic voltage and frequency scaling (DVFS)
of servers with low loads [33], and algorithms dependent on dynamic prediction
[29]. Each of those is affected by the type of load and client distribution change
that personal Cloud client request imposes. With consolidation-based solutions,
a more heterogeneous set of instances is favorable to counter challenges with
network bottlenecks and fault tolerance. DVFS methods could achieve more savings
1 The Era of the Personal Cloud: What Does It Mean for Cloud Providers? 7
by exploiting the personal Cloud instances with low activity. Moreover, energy
efficiency methods based on future request prediction could benefit from studying
personal Cloud load distribution and sizes.
Figure 1.4 summarizes the discussion on personal Cloud client behavior pre-
diction. The upper layer shows the defining elements of the client requests. This
includes knowledge about request source in terms of which client, from which
device, and from which location. These identifiers would affect the quality of service
conditions and can be used to offer customized and personalized service to clients.
Next, the application the clients access or use is central to the client behavior
modeling. The application type will affect VM instance on redundancy, network
bottlenecks and network resource configuration, and on application dependencies.
Such factors need to be taken into consideration. The volume of the requests is
the third aspect of this triangle. The request volume includes distributions of the
resource specification, lifetime or duration of each request, and the probabilistic
distribution that better represents arrival rate of requests. Considering the mentioned
aspects is critical for Cloud providers when planning the resource allocation
policies, energy efficiency policies, high availability policies, and service-pricing
policies.
8 M. Abu Sharkh et al.
To show how central Internet of Things implementation tiers are to the personal
Cloud, it is enough to look at the Internet Architecture Board (IAB) definition of IoT
in RFC 7452.33 “Architectural Considerations in Smart Object Networking”: “The
term ‘Internet of Things’ (IoT) denotes a trend where a large number of embedded
devices employ communication services offered by the Internet protocols. Many of
these devices, often called ‘smart objects’, are not directly operated by humans, but
exist as components in buildings or vehicles, or are spread out in the environment.”
These smart objects will represent the elements of the personal Cloud. The
personal Cloud of a specific user contains not only the traditional computa-
tion/communication devices (laptop, tablet, mobile phone), but it will be mostly
represented by other smart objects repressing other devices owned by the client like
the ones mentioned in the definition. This would range from fridges alerting clients
when grocery items are missing to sensors used to get insights on traffic to nodes
used in farms.
To further understand how this works, we look at the communication models
envisioned for the IoT and personal Cloud devices.
Devices can communicate in a device-to-device fashion where devices exchange
data between each other either through the Internet or other IP networks or through
direct connections using Bluetooth or ZigBee protocols, for example. Another type
of communication is device to Cloud [4, 9, 20, 31, 39, 50, 51].
“In a device-to-Cloud communication model, the IoT device connects directly
to an Internet Cloud service like an application service provider to exchange data
and control message traffic”[27]. This happens mostly through Wi-Fi connections.
The third model is device-to-gateway model. In the device-to-gateway model, the
IoT device connects through an application layer gateway service in order to reach
a Cloud service.
1.4.2.1 Volume
The sheer numbers of the IoT devices take the probability of risk to new levels.
This volume also means that a specific security vulnerability in one device will be
magnified as the tendency is to build and use high numbers of identical devices. A
poor design for a device or a vulnerability in a communication protocol could have
magnified impacts on the Cloud this device is dealing with.
10 M. Abu Sharkh et al.
The high volume of small cheap numerous devices means the manufacturers find
it harder to break even if they used more sophisticated processors or chips with more
memory. This limits the security functionalities/capabilities of a single thing.
1.4.2.2 Interconnectivity
The challenge here is the privacy of data collected by IoT devices is susceptible
for many reasons. First, the clients agreeing to data being collected is not as
straightforward as it was previously. The nature and frequency of data collected
or sent by the thing makes it harder for the client to get involved. Second, some IoT
devices do not commend an interface to configure privacy settings. Moreover, the
data collection and processing can always affect people and devices not involved
with the IoT ownership. Examples are people sharing the same geographic location
temporarily or sharing the same computing resources. Additional privacy protection
measures are required for every carrier of smart devices because as the time goes on
there is more of an assumption that each piece of data that can be collected will be
collected by some device around you [27, 49].
1 The Era of the Personal Cloud: What Does It Mean for Cloud Providers? 11
The notebook PC vs. tablet shipment and forecast report in [44] reveals a prime
driver behind the personal Cloud. The notebook PC share decreased from 75%
vs. 25% in 2011 to 27% vs. 73% for the tablet PC in 2015. Substantial changes
are happening. This does not account smartphones which at a fraction of the PC
weight are offering very similar video, audio, and processing capabilities to PCs
a few years ago. Alcatel-Lucent note the major factors affecting the evolution of
the mobile market and the close affiliation with the Cloud [6]. One factor cited by
Alcatel is the users’ aspiration to operate personal Clouds. This was paired with
the increasing demand for Wi-Fi and increasing demand for video conversation in
business environments.
The key take away from the chapter is that there is a wide space for improvement
for such a new paradigm and for how Cloud providers handle it. With the amount of
PC and tablet shipments in the world almost doubling from 2008 to 2015 reaching
545 million devices in 2015, there is definitely a space for growth [38].
The report in [45] gives us an insight into this potential. The report surveys the
population’s usage/interest in using public services through a Cloud medium. The
interest is really high for countries still migrating to the Cloud including India (81%
of the sampled population either using government service or interested in using
them), UAE (80%), and Brazil (79%). However, these numbers are significantly
lower for the UK, the USA, and Germany (59%, 55%, and 53%, respectively). There
is some way for the personal Cloud transformation to be complete in the minds
of the individual users. Providers need to investigate the areas where performance
or service offerings cause that gap in expectation and address these issues. One
attempt at figuring this out is in [43]. Within the set of users using the Cloud storage
services in the UK in 2014, 55% faced problems while storing or sharing files.
These problems included slow speed of access or use, technical server problems,
incompatibility and portability problem between devices or file formats, vendor
lock-in, privacy challenges with disclosure of data to third party, or unauthorized use
of information by service provider. The last one is an issue for the public with doubts
over agreements and government usage/ownership of data very much in the public’s
mind. Guarantees for data access and movement within the Cloud are expected by
the client. This is really consistent with the results from [41] which alluded that ease
of accessing files from several devices or locations was a main reason for using the
Cloud. “Among the reasons for using Cloud services, three in five users of Internet
storage space identified the possibility to use files from several devices or locations
and to share files with other persons easily (59%). Furthermore, over half of users
gave protection against data loss as a reason for using the Cloud (55%). Users also
valued having a larger memory space (44%) and being able to access large libraries
of music, films, or TV programs (22%).”
Concerns about security and privacy remain a main barrier to the use of Cloud
services.
For the cases where people were aware of Cloud services but chose not to use
them, “44% cited security or privacy concerns as a reason for not making use of such
services, 28% mentioned concerns about the reliability of service providers and 22%
cited a lack of skills” [41]. These insights draw a path for Cloud providers in order
to increase the personal Cloud penetration and achieve higher client satisfaction.
1 The Era of the Personal Cloud: What Does It Mean for Cloud Providers? 13
References
1. M. Abu Sharkh, A. Shami, An Evergreen Cloud: Optimizing Energy Efficiency in Green Cloud
Computing Environments Based on Virtual Machine State Prediction, Submitted to Elsevier
computer networks journal, Mar 2016.
2. M. Abu Sharkh, A. Shami, An evergreen cloud: optimizing energy efficiency in heterogeneous
cloud computing architectures. Veh. Commun. 9, 199–210 (2017)
3. M. Abu Sharkh, M. Jammal, A. Ouda, A. Shami, Resource allocation in a network-based cloud
computing environment: design challenges. Commun. Mag. IEEE 51(11), 46–52 (2013)
4. M. Abu Sharkh, A. Kanso, A. Shami, P. Öhlén, Building a cloud on Earth: a study of cloud
computing data center simulators. Comput. Netw. J. 108, 78–96 (2015)
5. M. Abu Sharkh, A. Shami, A. Ouda, Optimal and suboptimal resource allocation techniques in
Cloud computing data centers. J. Cloud Comput. Springer 6(1), 6 (2017) [Online]. Available:
https://link.springer.com/article/10.1186/s13677-017-0075-2
6. Alcatel Lucent, Mobile conversations in the era of the personal Cloud, Application note
[Online]. Available: http://www.tmcnet.com/tmc/whitepapers/documents/whitepapers/2013/
8057-alcatel-lucent-mobile-conversations-the-era-the-personal.pdf
7. Amazon, Amazon EC2 Pricing page [Online]. Available: https://aws.amazon.com/ec2/spot/
pricing/
8. A.Bankole Cloud client prediction models for cloud resource provisioning in a multitier web
application environment, An M.Sc. thesis Department of Systems and Computer Engineering,
Carleton University, Ottawa, Apr 2013
9. E. Baccarelli, P.G.V. Naranjo, M. Scarpiniti, M. Shojafar, J.H. Abawajy, Fog of everything:
energy-efficient networked computing architectures research challenges and a case study.
Access IEEE 5, 9882–9910 (2017). ISSN:2169-3536
10. A. Beloglazov, R. Buyya, Energy efficient resource management in virtualized Cloud data
centers, in 2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid
Computing
11. L.F. Bittencourt, M. Moraes, I. Petri, O.F. Rana, Towards virtual machine migration in fog
computing, in 2015 10th International Conference on P2P Parallel Grid Cloud and Internet
Computing (3PGCIC), 2015, pp. 1–8
12. F. Bonomi, R. Milito, P. Natarajan, J. Zhu, Fog computing: a platform for Internet of things and
analytics, in Big Data and Internet of Things: A Roadmap for Smart Environments (Springer,
Cham, 2014), pp. 169–186
13. D. Bruneo, A. Lhoas, F. Longo, A. Puliafito, Modeling and evaluation of energy policies in
green clouds. IEEE Trans. Parallel Distrib. Syst. 26(11), 3052–3065 (2015)
14. E. Caron, F. Desprez, A. Muresan, Forecasting for grid and cloud computing on-demand
resources based on pattern matching, in 2nd IEEE International Conference on Cloud
Computing Technology and Science, Nov 2010, pp 456–463
15. B.G. Chun et al., An energy case for hybrid datacenters. ACM SIGOPS Oper. Sys. Rev. 44(1),
76–80 (2010)
16. Cisco, Cisco global cloud index: forecast and methodology, 2014–2019 white paper. Avail-
able: http://www.cisco.com/c/en/us/solutions/collateral/service-provider/global-Cloud-index-
gci/Cloud_Index_White_Paper.html
17. L. Colombus, IDC Predicts SaaS Enterprise Applications Will Be A $50.8B Market By 2018
[Online]. Available: https://www.forbes.com/sites/louiscolumbus/2014/12/20/idc-predicts-
saas-enterprise-applications-will-be-a-50-8b-market-by-2018/#2657e46222a8. Accessed
Feb 2018
18. L. Columbus, Roundup of small and medium business cloud computing forecasts and
market estimates. [Online] Available: https://www.forbes.com/sites/louiscolumbus/2017/04/
29/roundup-of-cloud-computing-forecasts-2017/#1dfc95c731e8. Last accessed 27 Aug 2018
14 M. Abu Sharkh et al.
38. NPD DisplaySearch, Quarterly mobile PC shipment and forecast report (Jan 2013),
[Online]. Available: http://www.tmcnet.com/tmc/whitepapers/documents/whitepapers/2013/
8057-alcatel-lucent-mobile-conversations-the-era-the-personal.pdf
39. F. Al-Turjman, Fog-based caching in software-defined information-centric networks. Comput.
Electr. Eng. J. 69(1), 54–67 (2018). Elsevier
40. R. Samani, 3 key security challenges for the Internet of things [Online]. Available: http://www.
securingtomorrow.com/blog/knowledge/3-key-security-challenges-internet-things/
41. H. Seybert, P. Reinecke, Internet and Cloud services – statistics on the use by individuals: Half
of Europeans used the Internet on the go and a fifth saved files on Internet storage space in
2014 [Online]. Available: http://ec.europa.eu/eurostat/statistics-explained/index.php/Internet_
and_Cloud_services_-_statistics_on_the_use_by_individuals
42. S. Srikantaiah, A. Kansal, F. Zhao, Energy aware consolidation for Cloud computing, in
Proceedings of the 2008 Conference on Power Aware Computing and Systems, pp. 10–16
43. Statista Team, Problems experienced when saving or sharing files on the internet in Great
Britain in 2014 [Online]. Available: http://www.statista.com/statistics/321310/problems-with-
online-file-saving-or-sharing-great-britain-uk/
44. Statista Team, PC, tablet and Internet/Cloud-computing systems unit shipments worldwide
from 2002 to 2015 (in million units) [Online]. Available: http://www.statista.com/statistics/
282549/worldwide-pc-and-tablet-shipments/
45. Statista Team, Would you personally like to use Cloud computing in the future when
interacting with government departments offering public services? [Online]. Available: http://
www.statista.com/statistics/384604/Cloud-computing-public-service-usage/
46. Talking Cloud team, Cloud Services Brokerage Company List and FAQ [Online]. Avail-
able: http://talkinCloud.com/Cloud-services-broker/Cloud-services-brokerage-company-list-
and-faq
47. C. Tulkoff, Using immersion cooling technology for enhanced efficiency & reliability for
data center servers [Online]. Available: http://www.slideshare.net/CherylTulkoff/df-r-webinar-
immersion-cooling-technology. Accessed Oct 2015
48. K. Weins, Cloud Computing Trends: 2015 State of the Cloud Survey [Online].
Available: http://www.rightscale.com/blog/Cloud-industry-insights/Cloud-computing-trends-
2015-state-Cloud-survey. Accessed Oct 2015
49. Wind River, Security in the Internet of things lessons from the past for the connected
future [Online]. Available: http://www.windriver.com/whitepapers/security-in-the-internet-of-
things/wr_security-in-the-internet-of-things.pdf
50. W. Zhang, Z. Zhang, H. Chao, Cooperative fog computing for dealing with big data in the
internet of vehicles: architecture and hierarchical resource management. Commun. Mag. IEEE
55, 60–67 (2017). ISSN:0163-6804
51. J. Zhu, D.S. Chan, M.S. Prabhu, P. Natarajan, H. Hu, F. Bonomi, Improving web sites perfor-
mance using edge servers in fog computing architecture, in 2013 IEEE Seventh International
Symposium on Service-Oriented System Engineering, Redwood City, 2013, pp. 320–323.
https://doi.org/10.1109/SOSE.2013.73
Chapter 2
Optimization in Edge Computing
and Small-Cell Networks
2.1 Introduction
From the above discussion, the need for Edge Computing and Edge Communication
is clearly evident. We now focus our attention on one application as an example
which highlights finer requirements of the Edge Computing infrastructure.
Hospital is one of the applications which requires Edge Computing infrastructure
to deal with delay-sensitive patient monitoring system. It should be kept in mind
that critically ill patients effectively require real-time processing of the diagnostic
data and real-time intervention by the doctor. Some regular patients generate
diagnostic data that leads to predictable demands of edge resources (computa-
tion/communication). But some type of data generated by patients in unpredictable.
Hence, the Edge Computing paradigm leads to flexibility in provisioning edge
resources based on demand.
In most hospitals, various type of data (e.g., diagnostic data) with varied
processing time constraints are generated. For instance, the data has features of Big
Data. On the other hand, storage of medical records of past and present patients
needs to be done. The processing of such medical records to mine interesting
patterns could be done using Cloud Computing resources. Thus, Cloud Computing
and Edge Computing complement each other. Effectively real-time/non-real-time
processing of patient data effectively aids medical diagnosis.
The Edge Computing paradigm effectively has the following characteristics
needed by various applications.
2 Optimization in Edge Computing and Small-Cell Networks 21
This section briefly describes the efforts made by the researchers to explain the
need of Edge Computing. Further, the optimization strategies proposed for the edge
servers’ deployment would be discussed in this section. Amin et al. [4] discussed
the participatory Edge Computing for the local community services. They used
and proved that using the local server is better if we have a complex core network
available. They successfully implemented and provided third-party applications on
local servers for the community near to them. Kim et al. [5] worked with IoT devices
to offload their computation on the Cloud. Based on the results, authors found it
better to offload and compute on the Cloud rather than doing computation on the IoT
device itself. Samie et al. [6] described a way of distributed computation offloading
to many remote machines for distributed QoS for IoTs. It suggests to decompose
the problem into small processes among many resources. Dynamic programming
is used for resource allocation, and the whole problem is solved using ILP. Mao
et al. [7] went one step further and offloaded the computation data to the edge server
which is nearer as compared to the Cloud.
By many examples it has been proved by the researchers that even though the
machines on the edge have lesser configuration, still the delay is higher while getting
the computation results from the Cloud. It happens due to the communication
cost. In [8], the author proposed the optimizing solution by formulating it as an
22 J. Grover and R. M. Garimella
integer nonlinear program for offloading and resource allocation in Mobile Edge
Computing. The author found that due to the hardness of the problem, optimally
solving it for the last scale network is impractical. To solve it efficiently the problem
is divided into resource allocation and task offloading. Resource allocation problem
is further divided into two subproblems which are uplink power allocation and
computing resource allocation. Convex and quasi-convex optimization techniques
are used to solve these problems.
Now, this section briefly describes a few cases where researchers tried to optimize
the edge server deployment strategy. Qiang Fan et al. [9] gave a strategy CAPABLE
to optimize the cost of server deployment and end-to-end delay between client
devices and resources. The simulation results show that the scheme can trade off
between deployment cost and delay. Qin et al. [10] have proposed a software-
defined approach to manage heterogeneous IoT and sensor devices. This is done
via providing the best matching resource for different classes of IoT devices.
Authors have utilized a reflective middleware with a layered IoT SDN controller for
managing various IoT applications. It is the extension of multi-network information
architecture (MINA). Farah Slim et al. [11] proposes a mechanism related to the
multidimensional Cloud. It gives an analytical model for blocking analysis. It also
tried to find out the best strategy for distributed edge placement. Further, a strategy
is devised for resource allocation and capacity planning on the edge network. So
authors tried to optimize the limited resources available at the edge network.
For achieving energy efficiency after taking feedback latency into account,
mobile units demand for a common/different Cloud server for processing [12]. The
proposed algorithm gives a good performance for IoT with multi-small-cell Edge
Computing and MIMO. The problem is formulated as the minimization problem
but doesn’t use learning methods to optimize the device-edge combinations. The
proposed framework also works for radio access point implemented in distributed
and parallel manner with limited signaling to the Cloud. It is compared with disjoint
optimization algorithm and shows better results. In IoT-Edge computing scenario, it
is very much necessary to learn the best combinations over the period of time.
This section covers some of the related proposed solutions, but it is very clear that
very few researchers are focusing on methods for optimizing the allocation of edge
computation and communication resource to the IoT devices for their applications.
Further in the chapter, we have shown how Integer Linear Programming can be used
for optimal allocation of edge resources.
Consider devices/things which are locally distributed in space with limited local
computation/memory resources (e.g., nodes in the wireless sensor network). They
require high-end Edge Computing platforms. Also, it is necessary to be able to
2 Optimization in Edge Computing and Small-Cell Networks 23
schedule the computing tasks onto a relatively high-end Edge Computing server.
The goal is to minimize the cost as well as the delay in processing the tasks required
by local devices/things. We now provide the detailed modeling assumptions.
• There are two types {T ypeA, T ypeB} of Edge Computing devices with costs
{CA , CB } (model can easily be generalized to the finite number of types of high-
end computing devices).
• The devices/things are distributed in a rectangular grid (model can be generalized
to an arbitrary graph connectivity of devices)
The objective is to minimize the total cost of high-end computing platforms
(required to process the “delay sensitive” tasks submitted by the devices/things)
while at the same time ensuring that the things distributed on the rectangular
grid are able to schedule their tasks onto a certain minimum number of high-end
Edge Computing platforms. We now formulate the optimization problem as a {0, 1}
(binary/integer) linear programming problem.
Let the variable associated with placement of high-end Edge Computing plat-
forms on the rectangular grid be {aij , bij where 0 ≤ i ≤ M, 0 ≤ j ≤ N}, i.e.:
Let the grid points on the rectangular grid (where devices/things and edge
computers are located), i.e., (M)(N) points, be divided into mutually disjoint sets
{Dγ : 1 ≤ γ ≤ L}, i.e., those sets constitute a set partition of all points on the
rectangular grid. The constraints are that points in each set Dγ (for all j) are served
by at least “S” high-end Edge Computing platforms. Thus the optimization problem
has a linear objective function
M
N
M
N
CA aij + CB bij (2.3)
i=1 j =1 i=1 j =1
This integer ({0, 1}) linear programming problem is solved using well-known
techniques. Efficient algorithms exist for solving the problem.
An alternate formulation requires that every set Dδ (in the set partition of (M)(N)
grid points) is covered by at least S1 , type A computers and by at least S2 , type B
computers. Thus, the constraints in the above linear programming problem (Eq. 2.4)
get modified in the following manner:
aij ≥ S1 ;
(i,j )∈DK for 1 ≤ K ≤ L (2.5)
(i,j )∈DK bij ≥ S2 ;
We now discuss the high-end Edge Computing allocation scheme for different
IoT applications.
After the deployment of the edge servers, next issue that is required to be addressed
is allocation of different types of edge servers to different types of applications
according to their requirements. There can be many types of tasks taken care of
by the edge servers like image classification, audio/video classification, sensed data
fusion, data analytics, mathematical calculations for output generation, etc. These
tasks can be related to IoT applications such as healthcare, environment monitoring,
urban security, smart water/electricity management system, wearable, automated
transportation, smart surveillance, etc.
Now considering that there are two or more types of edge servers (high and low
cost) available on the edge, allocation of appropriate server to different applications
is necessary to complete them in the best possible ways. This can be done by
allocating cost A servers to one type of tasks and cost B servers to other type of
tasks and so on (depending upon the type of servers available). The division of tasks
can be majorly dependent upon the following:
• strict delay requirements of the task
• reliability requirement
• frequency of requests generated by an application and some other negligible
factors.
For example, image classification, smart surveillance, automated transportation
system, etc. related tasks should be assigned to an edge server with more powerful
resources, near to the source to reduce delay, with less probability to failure, and able
to manage frequent continuous requests. On the other hand, tasks like mathematical
calculations, function-based outputs, etc. can be given to the low-cost edge servers.
In the first case, delay more than threshold or failure in response can be catastrophic.
Based on this, optimization function can be updated with more constraints.
2 Optimization in Edge Computing and Small-Cell Networks 25
Expected Requests
P1
t12
Failure Probabilities
P2 a2 2 2 b2
P3 a3 3 3 b3
Pm am tmn bn
m n
Processing
Delav
n
minimize tij xij ∀ 1 ≤ i ≤ m
j =1
(2.6)
and
n
j=1 xij
minimize π m
i=1 pi
subject to constraints
n
xij ≤ ai ∀ i = 1, 2, . . . , m (2.7)
j =1
26 J. Grover and R. M. Garimella
m
xij = bi ∀ j = 1, 2, . . . , n (2.8)
i=1
Equation (2.6) is the updated objective function for resource allocation. This
multi-objective function is trying to minimize the cost of deployment, total time
taken for the processing of all the available requests, and the failure probability
associated with all the edge servers. Equation (2.7) keeps a check on the number
of requests to be sent to a server that should not exceed its capacity. Equation (2.8)
divides the load from an IoT node to edge servers. Equation (2.9) puts a lower bound
on the allocations. The following condition must be satisfied to achieve feasible
solution:
m
n
ai >= bj (2.10)
i=1 j =1
In case of not meeting the condition given in Eq. (2.10), more edge servers are
required to be added to satisfy the demand. The edge server can transfer the requests
to other neighbor edge servers as well in that situation. In worst scenario, if no such
servers are available, then the request can be transferred to Cloud as well.
It is expected that small cells (pico/femto), enabling efficient frequency reuse, will
be the important innovation in the deployment of the 5G cellular network. We
consider the case where the wireless communication network providing connectivity
between devices/things at the edge is infrastructure based.
The goal is to make efficient utilization of channels available in pico/femto cell.
Thus, we are naturally led to dynamic channel allocation based on historical traffic
data (at the edge) in adjacent small cells. Further, we propose dynamic spectrum
access based on time-optimal spectrum sensing (using CR approach). We formulate
these approaches as optimization problems. We reason below that these are two
related optimization problems. In [13], the time-optimal spectrum sensing problem
was formulated and solved. We utilize the solution for efficient spectrum sensing in
small cells. We now formulate and solve the dynamic channel allocation problem in
adjacent small cells and propose a solution.
Let there be M adjacent cells. Also, let the historical traffic (on some time unit
example hour) in the spectrum bands in those cells be {n1 , n2 , n3 , . . . , nM }. We
normalize ni s to arrive at the probability mass function, i.e.:
2 Optimization in Edge Computing and Small-Cell Networks 27
ni
pi = M ; for 1 ≤ i ≤ M (2.11)
j =1 nj
IDEA: In those bands where “pi ” is small, allocate small number of channels,
i.e., ni s. On the contrary if pi is allocated higher number of channels in such a way
that
M
ni .pi = E[Z] (2.12)
i=1
M
ni = L (2.13)
i=1
i.e., average number of channels allocated to a small cell are maximized. It should be
noted that “Z” is the random variable associated with number of channels allocated
per cell.
We reason below that, if there are no constraints imposed on ni s, the problem
becomes trivial.
Problem formulation: Maximize E[Z]
Subject to the constraint that
M
ni = L (2.14)
i=1
q1 ≤ q2 ≤ . . . ≤ qM (2.15)
Since {n1 , d} are integers, we are led to the linear Diophantine equation
Thus, we would like to utilize solutions of the above Eq. (2.17) and solve the
stochastic optimization problem of maximizing E[Z].
We first sort the probabilities, that is, pi s, in increasing order, resulting in
relabeled probabilities qi s, that is:
q1 ≤ q2 ≤ . . . ≤ qM (2.18)
n1 ≤
n2 ≤ . . . ≤
nM (2.19)
M
E[Z] =
n1 + (δ)d; where δ = (j − 1)qj (2.20)
j =1
M
V ar[Z] = (α − δ 2 )d 2 ; where α = (j − 1)2 qj (2.21)
j =1
The following theorem provides unique solution to the above optimization problem.
Theorem Assume that {(a1 , d1 ), . . . , (al , dl ), . . . , (ak , dk )} are the set of solutions
of linear Diophantine Eq. (2.17) among the infinitely many solutions, which are
positive real integers. If a1 < . . . < al < . . . < ak , then d1 > . . . > dl > . . . > dk .
In such a case, (a1 , d1 ) is the best solution, which maximizes the expected value
E(Y ).
Proof To prove that (a1 , d1 ) maximizes E(Y ), we need to prove the below
expression
al − a1
i.e. ≤δ for l = 2, 3, . . . , K (2.24)
d1 − dl
2 Optimization in Edge Computing and Small-Cell Networks 29
Since, (a1 , d1 ) and (al , dl ), both satisfy the linear Diophantine equation (2.17).
Therefore, we can write
(M − 1)M
M × a1 + × d1 = L (2.25)
2
(M − 1)M
M × a1 = L − × d1 (2.26)
2
(M − 1)M
M × al + × dl = L (2.27)
2
(M − 1)M
M × al = L − × dl (2.28)
2
Multiplying Eq. (2.24) by M on both sides, we get
al − a1
M× ≤M ×δ (2.29)
d1 − dl
M × al − M × a1
≤M ×δ (2.30)
(d1 − dl )
(M−1)M (M−1)M
[L − × dl ] − [L − × d1 ]
2 2
≤ M ×δ (2.31)
(d1 − dl )
(M−1)M
× (d1 − dl )
2
≤M ×δ (2.32)
(d1 − dl )
(M − 1)
≤δ (2.33)
2
Substituting the value of δ, we get
(M − 1)
M
≤ (j − 1)pj (2.34)
2
j =1
M
1
M
(M − 1)
(j − 1)pj = (j − 1) = (2.35)
M 2
j =1 j =1
(M − 1)
M
< (j − 1)pj (2.36)
2
j =1
Proof To prove the minimum value of M j =1 (j − 1)pj occurs only for uniform
distribution, i.e., p1 = p2 = . . . = pM = M 1
, we use the proof of contradiction.
Suppose that the probability masses p1 = q − , p2 = p3 = . . . = pM−1 = q = M1
and pM = q + gives the minimum value of M j =1 (j − 1)pj , where > 0 is a
small positive real
number. Using the given probabilities, we can calculate value of
the expression M j =1 (j − 1)pj as below:
M
M
(j − 1)pj = 0 × (q − ) + (j − 1) × q
j =1 j =1 (2.37)
+ (M − 1) × (q + )
M
M
(j − 1)pj = (j − 1) × q + (M − 1) × (q + )
j =1 j =1
M
= (j − 1) × q + (M − 1) ×
j =1
M (2.38)
1
= × (j − 1) + (M − 1) ×
M
j =1
1 (M − 1)M
= × + (M − 1) ×
M 2
(M − 1)
= + (M − 1) ×
2
Since,
> 0, it contradicts our supposition. Therefore, we can conclude that the
value of M j =1 (j − 1)pj >
(M−1)
2 for any probability distribution and minimum
value occurs for uniform distribution.
2 Optimization in Edge Computing and Small-Cell Networks 31
2.6 Conclusion
In this book chapter, we have solved the cost-effective edge server placement
problem optimally, considering the location of the edge servers are grid points.
This is solved using two types of edge servers that can easily be generalized to
any number of servers’ type. Efficient allocation of resources is also taken into
consideration, and this issue is also solved using another optimization function.
Along with this, we also solved the issue of optimal spectrum sensing arising in
cellular cognitive radio networks. This book chapter also solves the issue of optimal
dynamic channel allocation problem arising in small cells based on historical traffic
data. In future, cost effective edge placement optimization problem can be solved
based on a graph where the location of the edge servers can be arbitrary.
References
1. M. Satyanarayanan, P. Bahl, R. Caceres, N. Davies, The case for VM-based cloudlets in mobile
computing. IEEE Pervasive Comput. 8(4), 14–23 (2009)
2. M. Satyanarayanan, The emergence of edge computing. Computer 50(1), 30–39 (2017)
3. S. Sardellitti, G. Scutari, S. Barbarossa, Joint optimization of radio and computational
resources for multicell mobile-edge computing. IEEE Trans. Signal Inf. Process. over Netw.
1(2), 89–103 (2015)
4. A.M. Khan, F. Freitag, On participatory service provision at the network edge with community
home gateways. Proc. Comput. Sci. 109, 311–318 (2017)
5. S. Kim, Nested game-based computation offloading scheme for mobile cloud iot systems.
EURASIP J. Wirel. Commun. Netw. 2015(1), 229 (2015)
6. F. Samie, V. Tsoutsouras, S. Xydis, L. Bauer, D. Soudris, J. Henkel. Distributed QoS
management for internet of things under resource constraints, in 2016 International Conference
on Hardware/Software Codesign and System Synthesis (CODES+ ISSS) (IEEE, 2016), pp. 1–
10
7. Y. Mao, J. Zhang, K.B. Letaief, Dynamic computation offloading for mobile-edge computing
with energy harvesting devices. IEEE J. Sel. Areas Commun. 34(12), 3590–3605 (2016)
8. T.X. Tran, & D. Pompili, Joint task offloading and resource allocation for multi-server mobile-
edge computing networks. arXiv preprint arXiv:1705.00704 (2017)
9. Q. Fan, N. Ansari. Cost aware cloudlet placement for big data processing at the edge, in 2017
IEEE International Conference on Communications (ICC), May 2017, pp. 1–6
10. Z. Qin, G. Denker, C. Giannelli, P. Bellavista, N. Venkatasubramanian. A software defined
networking architecture for the internet-of-things, in Network Operations and Management
Symposium (NOMS), 2014 IEEE (IEEE, 2014), pp. 1–9
11. F. Slim, F. Guillemin, Y. Hadjadj-Aoul, On virtual network functions’ placement in future
distributed edge cloud, in 2017 IEEE 6th International Conference on Cloud Networking
(CloudNet), Sept 2017, pp. 1–4
12. S. Sardellitti, G. Scutari, S. Barbarossa, Joint optimization of radio and computational
resources for multicell mobile-edge computing. IEEE Trans. Signal Inf. Process. Over Netw.
1(2), 89–103 (2015)
13. G.R. Murthy, R.P. Singh, S. Abhijeet, S. Chandhary, Time optimal spectrum sensing. arXiv
preprint arXiv:1606.02849 (2016)
Chapter 3
A Comprehensive Survey on Architecture
for Big Data Processing in Mobile Edge
Computing Environments
3.1 Introduction
Mobile networks have evolved significantly over the past few decades from 2G to
3G and further from 3G to 4G LTE. This growing mobile traffic is driven by con-
sumer smart phones, streaming video messaging, and P2P applications. In addition,
this growth is expected to increase more as businesses extend their processes to
smart mobile devices after the evolution of mobile cloud computing and services
such as m-learning, m-gaming, m-healthcare, m-commerce, m-governance, etc., are
directly accessible from the mobile devices. This also implies that large amount of
data is involved and high network load and growing demand of network bandwidth
as data have to be transmitted and received to and from the mobile devices and
cloud data centers [1]. With existing centralized cloud computing architecture, the
applications are facing harsh challenges. For example, mobile devices connected
to the distant centralized cloud servers try to obtain high computing applications
which enforce additional load both on the radio access networks and on the backhaul
networks and, thus, introducing high latency [2]. As estimated by Cisco, the number
of devices connected to the IoT will be around 50 billion by 2020, and at the same
time, mobile data traffic is predicted to continue doubling each year [3, 4]. However,
the load if this large volume of data which is generated by end users and processed
at cloud data centers leads to certain challenges like stringent latency, capacity
constraints, resource-constrained devices, uninterrupted services with intermittent
connectivity, and enhanced security [5].
There are two emerging paradigms which have been proposed to help overcome
these challenges: (i) cloud-based radio access network (CRAN) that aims at the
M. J. Kaur ()
Department of Engineering, Amity University Dubai, Dubai International Academic City, Dubai,
UAE
e-mail: [email protected]
centralization of base station (BS) functions via virtualization and (ii) mobile
edge computing (MEC) which proposes to empower the network edge [6]. CRAN
brings cloud computing technologies into mobile networks by centralizing baseband
processing units (BBU) of radio access network to ensure highly efficient network
operation and flexible service delivery when handling mobile internet traffic surging
[7]. CRAN uses centralized BBU to do baseband processing, while MEC handles
distributed task offloading by shifting computation capacity from a public cloud to
an edge cloud, which significantly reduces latency. In MEC, network edge devices,
such as base stations (BSs), access points, and routers, are endowed with, albeit
limited, computing and storage capabilities to serve users’ requests as a substitute
of clouds, while significantly reducing the transmission latency as they are placed
in the proximity of end users.
The first real-world MEC platform was introduced by Nokia Networks in 2013. In
that, the computing platform called radio applications cloud servers (RACS) – was
fully integrated with the flexi multiradio base station [19]. Later the term, mobile
edge computing was standardized by European Telecommunications Standards
Institute (ETSI) and Industry Specifications Group (ISG). It is acknowledged
by European 5G PPP (5G Infrastructure Public Private Partnership) as a prime
emerging technology for 5G networks [8].
The ETSI Mobile Edge Computing Industry Specification Group (MEC ISG)
was formed in December 2014. Its aim was “to create a standardized, open
environment which will allow the efficient and seamless integration of applications
from vendors, service providers, and third-parties across multi-vendor mobile edge
computing platforms” [9]. The ISG recently received a 2-year extension – beginning
in March 2017. The ISG modified its terms of reference to include the investigation
of the use of edge computing beyond a “mobile-only” access environment such as
areas with Wi-Fi and fixed access technologies. Therefore, ISG will be changing its
name from “mobile edge computing” to “multiaccess edge computing” to reflect the
change in terms of reference, while still maintaining the widely recognized “MEC”
acronym [10]. 3GPP has also recently published their work plan toward 5G and
more specifically toward meeting IMT-2020 requirements, which includes study
items and development of technical reports on scenarios that drive 5G architecture
and interfaces. Major standardization bodies, ITU (International Telecommunica-
tion Union), and European Telecommunications Standards Institute (ETSI) have
initiated activities relating to 5G [11–14] with commercial deployments expected
in 2020.
MEC enables cloud computing capabilities at the edge of the cellular network.
It also enables the applications and related processing tasks like content caching
and media processing to run on the edge of the network, i.e., more closer to the
mobile subscribers. This not only improves the user experience by providing high
3 A Comprehensive Survey on Architecture for Big Data Processing. . . 35
availability of the content with reduced latency, but also puts control of the service
delivery in the hands of mobile network operator (MNO). This further allows the
MNO to manage user experience and also helps to rapidly deploy new applications
and services to the subscribers [15]. MEC can be implemented within the existing
3GPP mobile network infrastructure. Therefore, it is expected to be integral part of
5G specifications.
The cloud service platforms will be providing the computation and other services
to the mobile communication networks with the help of huge alternative architecture
centralized data centers which are the backbone of the network. A signal or data
packet usually spends RTT (round trip time) of several milliseconds to travel
between the mobile device and the cloud in order to be processed in the data center
which might be several kilometers away. To reduce the RTT to the range of less than
1 ms, the distance between the mobile and cloud can be no greater than 100 km. But
the closest places or “hops” to a user are the base stations at the edge of the network
which can be used to outsource the cloud closer to the user providing low latency.
Therefore, in response to the above requirement, the mobile operators are working
on mobile edge computing (MEC) in which the computing, storage, and networking
resources are integrated with the base stations. Hence, all compute intensive and
latency sensitive applications like augmented reality and image processing can
be hosted at the edge of the network. This provides the end user with swift and
powerful computing, energy efficiency, storage capacity, mobility, location, and
context awareness support [16, 17]. A decentralized architecture with compute and
storage capability at the edge of the network provides an intermediate processing
stage to reduce the amount of data shipped back to the cloud by executing algorithms
for applications such as face recognition, load building, and landmark labeling
for augmented reality (AR) – a cognitive assistance or even cloud-sourced video
analytics. The communications service providers (CSPs) will need to respond fast
to capacity increases while maintaining latency.
According to the white paper published by ETSI, mobile edge computing can be
characterized by [18]:
• On Premises: MEC platforms can run isolated from the rest of the network,
while they have access to local resources. This is very important for machine-
to-machine scenarios. The MEC property of segregation from other networks
also makes it less vulnerable.
• Proximity: Being deployed at the nearest location, mobile edge computing has an
advantage to analyze and materialize big data. It is also beneficial for compute-
hungry devices such as segmented reality, video analytics, etc.
• Lower latency: Mobile edge computing services are deployed at the nearest
location to user devices, isolating network data movement from the core network.
Hence, user experience is accounted high quality with ultra-low latency and high
bandwidth.
• Location awareness: Edge-distributed devices utilize low-level signaling for
information sharing. MEC receives information from edge devices within the
local access network to discover the location of services.
36 M. J. Kaur
3.2.1 Framework
A general view of basic cellular network is shown in Fig. 3.1. The core network is
wire connected (e.g., IP/ethernet) with radio access network (RAN). RAN connects
the base station with backhaul network through the interface that supports high data
transfer rate [19]. Wireless user equipment connects through RAN to the mobile
operator network.
RAN facilitates the connection between mobile phones and the mobile core
network. It also covers the wide geographical area divided into several cells, and
each cell is integrated with its base station. Base stations are further connected to
the base station controller (BSC) via microwave or WLL. It is also known as radio
network controller (RNC) which is responsible to control the base station node and
and network level. At the mobile edge system level, the mobile edge orchestrator
serves the central role of coordinating among the UEs, the mobile edge hosts,
and the network operator. It records accounting and topological information about
the deployed mobile edge hosts, available resources, and available mobile edge
services. Also, it interfaces with the virtualization infrastructure and maintains
authentication and validation of application packages. It maintains information
on the whole mobile edge system about all the deployed hosts, the services and
resources available in each host, the applications that are instantiated, and the
topology of the network. It manages the applications by checking the integrity
and authenticity of the application, validating the policies for the applications, and
maintaining a catalog of the applications that are available.
At the mobile edge host level, mobile edge applications run VMs supported by
virtualization infrastructure within the mobile edge host. They provide services
like computational job execution and UE location information. The virtualization
infrastructure includes a data plane that executes the forwarding rules received by
the mobile edge platform and routes the traffic among the applications, services,
and the networks. It also manages the virtualized resources for the mobile edge
applications, which consists of allocating and releasing virtualized computation,
3 A Comprehensive Survey on Architecture for Big Data Processing. . . 39
storage, and network resources. It provides support for fault and performance
monitoring by collecting and reporting information on virtualized resources and
providing information further to server- and system-level management.
The mobile edge platform hosts mobile edge services. It interacts with mobile
edge applications, so that they can advertise, discover, offer, and consume mobile
edge services. The mobile edge platform manager provides element management
functions to the mobile edge platform and administers application essentials such
as life cycle, service requirements, operational rules, domain name system (DNS)
configuration, and security. In addition, mobile edge host can provide persistent
storage and time of the day information for the applications. The mobile edge
platform receives the traffic forwarding rules from the mobile edge platform
manager, mobile edge applications, and mobile edge services and based on those as
well as on policies it provides the instructions to the forwarding plane. Mobile edge
platform can communicate with other mobile edge platforms over reference point,
which is intended for control plane procedures. Using this interface, platforms may
be grouped together and can form a communications grid.
IoT devices are expected to generate a significant amount of data as their use
becomes ubiquitous. IoT cloud communication models and big data generated
by devices result in increased latency and incremental data transfer cost [24–26].
Using cloud computing with the IoT will raise concerns about privacy, and also
the most of the IoT end nodes are power constrained. While design considerations
are intended to improve the performance, the processing and memory capabilities
will be constrained. Offloading some computing tasks to the network edge could
be the solution – called mobile edge computing. MEC can address these issues by
serving as an aggregation point. MEC consists of geodistributed servers or virtual
servers with built-in IT services. These can be implemented locally at mobile user
premises. MEC may utilize cellular network elements like base station, Wi-Fi access
point, LTE base station, or a 3G/LTE cell aggregation site which can be indoor or
outdoor. Therefore, it works on both downstream data on behalf of cloud services
and upstream data on behalf of IoT services.
3.3.1 Applications
MEC offers high bandwidth and low latency mobile access to both information
and computation resources, because of its proximity to the mobile devices. Hence,
reducing the resource demands on the mobile backhaul. The mobile edge host is
40 M. J. Kaur
cloud server, and they can immediately diagnose the patients and can assist them
accordingly [30–34].
• Moving IoT devices and connected vehicles – The motivation behind the IoT
use cases is being connected to the network and also mobility with connectivity,
for example, drones, vehicles, etc. The connection scenario can be vehicle to
vehicle, vehicle to access point, or access point to access point. MEC offers local
processing and low latency to such situations and is expected to be important to
5G services. In fact, mobility and session management is usually performed by
the core network functions, for example, serving gateway (SGW) and serving
GPRS support node (SGSN) in the current 3G/4G networks. Mobile edge
applications for IoT would not require complex traffic filtering and manipulation
in order to receive the desired traffic and the mobile edge platform can leverage
on legacy network entities to carry out consistent control plane operations,
e.g., for mobility as well as other crucial aspects associated to gating, QoS
enforcement, charging, etc. Therefore, MEC enables scalable, reliable, and
distributed environments that are synced with the local sensors [35].
3.3.2 Challenges
In this section, we will discuss the challenges for deployment and implementation
of real-time data processing at the edge of the network [36, 37]. Current cloud
computing frameworks like Amazon Web Service, Microsoft Azure, and Google
App Engine can support data-intensive applications but data processing [38–40].
• Threats related to infrastructure – The concept of the edge network computing
is part of the “last mile network” which means that it uses many technologies
to build a network. While same is advantageous, and at the same time prone
to several types of attacks like DoS (Denial of Service) attacks and jamming.
These two types of attacks can easily consume bandwidth, frequency band, and
computing resources at the edge. Also, there could be rogue users that can intrude
the system and perform some malicious activities. For example, a device can be
reconfigured and set to send fake information or any incorrect data. Hence, any
compromised device connected in a cluster environment can change and control
services in that cluster.
• Virtualization – There are various challenges related to virtualization technol-
ogy – which is based to share the resources in the mobile edge computing
environment. Therefore, if one resource is compromised, it can affect the whole
virtualization infrastructure. One of the common security concerns is privacy
leakage. The APIs implemented to deliver the information should be protected
against any kind of malicious activities. The network functions virtualization
(NFV) is a core enabling technology employed by the mobile edge hosts to
create network elements like routers, packet gateways, and internet protocol (IP)
multimedia subsystems, using generic hardware. By separating software from
42 M. J. Kaur
Big data processing systems in mobile environments require ultra-low latency and
reliable data analytics solutions that can combine in real-time heterogeneous mix
of data stemming from the IoT mobile devices. Such data analytics capabilities
cannot be provided by cloud-centric data processing techniques. With higher
capability of smart mobile devices, a distributed mobile edge computing architecture
can leverage deep learning techniques for reliable mobile sensing, automate the
detection process, and also to detect the security threats. In this section, we will
survey the state-of-the-art research efforts on the same issue in the context of MEC.
Deep learning is a promising approach for extracting accurate information from
IoT devices deployed in complex environments. Deep learning is more suitable
for mobile edge computing due to its multilayer structure. Deep neural networks
3 A Comprehensive Survey on Architecture for Big Data Processing. . . 43
provide highly accurate classifiers which consist of different layers that follow each
other. Each layer is a simple function of its previous layer, representing a more
sophisticated concept than the previous layer. Therefore, the initial layer is the raw
input data, and the final layer gives the result of the inference.
Machine learning methods process the data streams by performing supervised,
unsupervised, semisupervised, and deep learning models [44]. The data mining
and machine learning methods are useful for early knowledge discovery from big
data streams. Therefore, these methods are useful for real-time big data analytics
where multiple learning models at different levels of big data systems filter the
data streams and uncover the knowledge patterns in parallel [45–47]. Deep neural
network (DNN) is the technology behind voice-based personal assistance [48],
self-driving cars [49], automatic image processing [50], etc., and finance analytics
has quickly adopted machine learning to harness large volume of data in areas of
fraud detection, risk management, and compliance. Various smart services are being
pushed to edge devices. For example, Intel’s Movidius Neural Compute Stick [51]
is a tiny deep learning device that one can use to accelerate AI programming and
DNN inference application deployment at the edge. Basically, DNNs are the types
of artificial neural networks (ANN) that work as function approximators used to
analyze the correlation between relevant features and the output of the processed
data. Therefore, DNNs are ANN with multiple hidden layers [52].
Deep neural networks (DNNs) can model high-level abstractions in data due to
feature hierarchy that allows each hidden layer to train on a distinct feature set
based on the previous hidden layer’s output. Hence, DNNs are able to recognize
and model complex features from large datasets like video, radar readings from
various IoT mobile devices. Various machine learning algorithms can be employed
for edge analytics in mobile environment like recurrent neural networks (RNNs),
convolutional neural networks (CNNs), restricted Boltzmann machines (RBMs),
generative adversarial networks (GANs), and long short-term memory (LSTM) [53–
56]. In this particular case of mobile sensing, recurrent neural networks (RNNs) are
more suitable for variable length sequential data like the one produced by mobile
sensors [57, 58]. Some other deep learning and the related issues are depicted in
Table 3.1.
In this section, we investigate the overall data processing workflow in collection and
analysis of broadcast messages from the base stations in the mobile environment.
These broadcast messages are continuously being transmitted from the base stations
44 M. J. Kaur
Table 3.1 Deep learning methods, description, and issues for mobile sensing
S.no. Method Description and issues
1 Residual nets [59] These introduce shortcut connections into CNNs, which greatly
reduces the difficulty of training super deep models. However,
since residual nets mainly focus on visual inputs, they lose the
capability to model temporal relationships, which are great
importance in time-series sensor inputs.
2 LRCN [60] These apply CNNs to extract features for each video frame and
combine video frame sequences with LSTM [61], which exploits
spatial-temporal relationships between video inputs. However, it
does not consider modeling multimodal inputs. This capability is
important to mobile sensing and computing tasks.
3 DBM [62] Multimodal DBMs merge multimodal inputs such as images and
text, with deep Boltzmann machines (DBMs). However, the work
does not model temporal relationships and does not apply tailored
structures such as CNNs to effectively and efficiently exploit
local interactions within input data.
4 RBM [63] and These use deep Boltzmann machines and multimodal DBMs to
MultiRBM [64] improve the performance of heterogeneous human activity
recognition.
5 IDNet [65] It applies CNNs to the biometric gait analysis task.
6 DeepX [66] and These reduce the energy consumption of deep neural networks,
RedEye [67] based on software and hardware, respectively.
LRCN long-term recurrent convolutional networks, LSTM long short-term memory, DBM deep
Boltzmann machines, RBM restricted Boltzmann machines, RNN recurrent neural networks, CNN
convolutional neural networks
and dictate how mobile phones interact with them. These messages also include
information on how the base station communicates with other base stations and
handles cellular traffic in the surrounding RF environment. The broadcast messages
are decoded, and the resultant textual data are processed and used as input to
the machine learning algorithms [68]. LTE technology, which is based on UMTS,
is optimized for all-IP traffic and has several advantages including high network
throughput and low latency.
There are many existing methods to detect the base station routers (BSRs) that
use global positioning system (GPS) for location information. These methods are
time lapse based or configuration based. However, in case of mobile edge com-
puting, flexible and more efficient solutions are required, which can automatically
detect the presence of noncommercial BSRs without the prior need of location
information, extensive time-lapse surveys or without scanning the databases of
known commercial carriers so as to reduce the comeback time. Hence, there is a
need for automated data mining techniques to understand the various patterns in
big data generated by the mobiles and base stations. Most of the state-of-the-art
methods focus on unsupervised clustering which is the most popular data mining
method used for finding hidden patterns in data. Table 3.2 gives a brief on some of
machine learning methods.
3 A Comprehensive Survey on Architecture for Big Data Processing. . . 45
The ubiquity of mobile phones and the associated big data consist of large and
complex datasets generated by the wide variety of applications and other sensor
data involved. With the implementation of MEC near the mobile devices can help
the data analytics extracting the meaningful information from the raw data on
the mobile edge node which could be then sent to the cloud servers for further
processing. This will reduce the computational load on the cloud server for storage,
data sync, and sharing. Deployment of MEC can solve many problems involving
bandwidth, battery life, and storage to the resource-constrained mobile devices. In
this chapter, we had a brief discussion on how the machine learning methods can
help tailor the needs of the mobile edge analytics to overcome various challenges
of mobile sensing, detection of the base station routers, and also the detection
of attacks for secure and reliable solutions. For the future work, it would be
interesting to investigate other deep learning algorithms for big data processing in
MEC environment and apply these algorithms on different datasets to examine their
effectiveness. Comparison of performance of different canonical approaches can be
done, and potential applications of more sophisticated deep learning techniques can
be explored. MEC still being in early stages of research and development have many
challenges and, at the same time, give a variety of innovative opportunities in the
future.
References
11. ITU, Work Plan, timeline, process and deliverables for the future development of IMT (2015)
12. ITU, High level 5G architecture. Network softwarization, gaps (2013). http://www.itu.int/en/
ITU-T/focusgroups/imt-2020/Pages/default.aspx
13. ETSI, 5G – the 5G infrastructure public private partnership: the next generation of communi-
cation networks and services (2015)
14. ETSI, Mobile edge computing – introductory technical whitepaper (2014)
15. Intel, Real-world impact of mobile edge computing. White paper (2016)
16. S. Yi, C. Li, Q. Li, A survey of fog computing: concepts, applications and issues, in
Proceedings of the 2015 Workshop on Mobile Big Data, ACM, New York, 2015. pp. 37–42
17. Y. Jararweh, A. Doulat, A. Darabseh, M. Alsmirat, M. Al-Ayyoub, E. Benkhelifa, Sdmec:
software defined system for mobile edge computing, in IEEE International Conference on
Cloud Engineering Workshop (IC2EW), April 2016, pp. 88–93
18. M. Patel, B. Naughton, C. Chan, N. Sprecher, S. Abeta, A. Neal, Mobile edge computing
introductory technical white paper. White paper. Mobile Edge Computing (MEC) industry
initiative (2014)
19. C.K. Park, Performance for radio access network in mobile backhaul network. J. Inst. Internet.
Broadcast. Commun. 12(6), 297–302 (2012)
20. A.H. Khan, M.A. Qadeer, J.A. Ansari, S.Waheed, 4G as a next generation wireless network, in
Future Computer and Communication ICFCC 2009, 2009, pp. 334–338
21. CommVerge, Radio Access Network (RAN) Optimization (2016). Available: http://
www.commverge.com/Solutions/SubscribersServicesManagement/RANOptimization/tabid/
174/Default.aspx
22. J. Wu, Z. Zhang, Y. Hong, Y. Wen, Cloud radio access network (C-RAN): a primer. IEEE Netw.
29(1), 35–41 (2015)
23. Y. Liu, M.J. Lee, Y. Zheng, Adaptive multi-resource allocation for cloudlet-based mobile cloud
computing system. IEEE Trans. Mob. Comput. 15(10), 2398–2410 (2016)
24. F. Akhbar, V. Chang, Y. Yao, V.M. Munoz, Outlook on moving of computing services towards
the data sources. Int. J. Inf. Manag. 36, 645–652 (2016)
25. C.S. Li, F. Darema, V. Chang, Distributed behavior model orchestration in cognitive internet
of things solution. Enterp. Inf. Syst. (2017). https://doi.org/10.1080/17517575.2017.1355984
26. M. Mital, V. Chang, P. Choudhary, A. Pani, Z. Sun, Adoption of cloud based inter-
net of things in India: a multiple theory perspective. Int. J. Inf. Manag. (2016).
https://doi.org/10.1016/j.ijinfomgt.2016.02.011
27. International Data Corporation, IDC Reveals Worldwide Internet of Things Predictions for
2015. Press Release (2015)
28. K. Hong, D. Lillethum, U. Ramachandran, B. Ottenwalder, B. Koldehofe, Mobile fog: a
programming model for large-scale applications on the internet of things, in Proceedings of
the Second ACM SIGCOMM Workshop on Mobile Cloud Computing MCC’13, New York, NY,
USA, 2013, pp. 15–20. [Online]. Available: http://doi.acm.org/10.1145/2491266.2491270
29. International I, Computer technology increasingly aids traffic management (2009). http://
www.itsinternational.com/categories/detection-monitoring-machine-vision/features/
30. A.V. Dastjerdi, H. Gupta, R.N. Calheiros, S.K. Ghosh, R. Buyya, Fog computing: principles,
architectures and applications. arXiv preprint arXiv:1601.02752 (2016)
31. Y. Cao, S. Chen, P. Hou, D. Brown, Fast: a fog computing assisted distributed analytics
system to monitor fall for stroke mitigation, in IEEE International Conference on Network,
Architecture and Storage (NAS), 2015, pp. 2–11
32. A.S. Go, D. Mozaffarian, V.L. Roger, E.J. Benjamin, J.D. Berry, M.J. Blaha, S. Dai, E.S. Ford,
C.S. Fox, S. Franco, et al., Heart disease and stroke statistics – 2014 update. Circulation 129(3)
(2014)
33. P.A. Heidenreich, J.G. Trogdon, O.A. Khavjou, J. Butler, K. Dracup, M.D. Ezekowitz, E.A.
Finkelstein, Y. Hong, S.C. Johnston, A. Khera, et al., Forecasting the future of cardiovascular
disease in the United States a policy statement from the American Heart Association.
Circulation 123(8), 933–944 (2011)
48 M. J. Kaur
34. V. Stantchev, A. Barnawi, S. Ghulam, J. Schubert, G. Tamm, Smart items, fog and cloud
computing as enablers of servitization in healthcare. Sens. Transducers 185(2), 121 (2015)
35. S. Datta, C. Bonnet, Haerri, Fog computing architecture to enable consumer centric internet of
things services, in IEEE International Symposium on Consumer Electronics (ISCE), pp. 1–2
36. W. Li, Y. Zhao, S. Lu, D. Chen, Mechanisms and challenges on mobility- augmented service
provisioning for mobile cloud computing. IEEE Commun. Mag. 53(2), 978–995 (2015)
37. H. Hromic, D. Le Phuoc, M. Seerano, A. Antonic, I.P. Zarko, C. Hayes, S. Decker, Real time
analysis of sensor data for the internet of things by means of clustering and event processing,
in Proceedings of the IEEE International Conference on Communications, 2015, pp. 685–691
38. http://aws.amazon.com
39. https://azure.microsoft.com/en-gb/
40. https://appengine.google.com/
41. B. Han, V. Gopalakrishnan, L. Ji, S. Lee, Network function virtualization: challenges and
opportunities for innovations. IEEE Commun. Mag. 53(2), 90–97 (2015)
42. A. Manzalini, R. Minerva, F. Callegati, W. Cerroni, A. Campi, Clouds of virtual machines on
edge networks. IEEE Commun. Mag. 51(2), 148–155 (2013)
43. C. Clark, K. Fraser, S. Hand, G. Hansen, E. Jul, C. Limpach, I. Pratt, A. Warfield, Live
migration of virtual machines, in Proceedings of the USENIX Symposium on Networked
Systems Design and Implementation (NSDI), Berkley, USA, 2005
44. M. Habib ur Rehman, P.P. Jayaraman, S.R. Malik, M. Medhat Gaber, RedEdge: a novel
architecture for big data processing in mobile edge computing environments. J. Sens. Actuator
Netw. 6, 17 (2017)
45. C.K.S. Leung, R.K. MacKinnon, F. Jiang, Reducing the search space for big data mining
for interesting patterns form uncertain data, in IEEE International Conference on Big Data,
Washington, DC, USA, 2014, pp. 315–322
46. A. Stateczny, M. Wlodarczyk-Sielicka, Self-organizing artificial neural networks into hydro-
graphic big data reduction process, in Rough Sets and Intelligent Systems Paradigms, Spirnger,
Berlin Germany, 2014, pp. 335–342
47. A. Ragyanszki, K.Z. Gerlei, A. Suranyi, A. Kelemen, S.J.K. Jensen, I.G. Csizmadia, B.
Viskolcz, Big data reduction by fitting mathematical functions: a search for appropriate
functions to fit Ramachandran surfaces. Chem. Phys. Lett. 625, 91–97 (2015)
48. M. Schuster, Speech recognition for mobile devices at Google, in Pacific Rim International
Conference on Artificial Intelligence, Springer, 2010, pp. 8–10
49. M. Bojarski, D. Del Testa, D. Dworakowski, B. Firner, B. Flepp, P. Goyal, L.D. Jackel, M.
Monfort, U. Muller, J. Zhang, et al, End to end learning for self- driving cars. arXiv preprint
arXiv:1604.07316 (2016)
50. Google, Google cloud vision api (2017). https://cloud.google.com/vision. Accessed 11 Nov
2017
51. Movidius, Movidius neural compute stick (2017). https://developer.movidius.com/. Accessed
11 Nov 2017
52. M. Chen, U. Challita, W. Saad, C. Yin, M. Debbah, Machine learning for wireless networks
with artificial intelligence: a tutorial on neural networks. arXiv preprint arXiv:1710.02913
(2017)
53. I. Goodfellow, Y. Bengio, A. Courville, Deep Learning (MIT Press, Cambridge, MA, 2016)
54. K. He, X. Zhang, S. Ren, J. Sun, Delving deep into rectifiers: surpassing human-level
performance on ImageNet classification, in Proceedings of the IEEE International Conference
on Computer Vision (ICCV), 2015, pp. 1026–1034
55. A. Krizhevsky, I. Sutskever, G. Hinton, ImageNet classification with deep convolutional neural
networks. Adv. Neural Inf. Proces. Syst. (NIPS) 25, 1097–1105 (2012). Curran Associates Inc
56. K. Katevas, I. Leontiadis, M. Pielot, J. Serra, Practical processing of mobile sensor data for
continual deep learning predictions, in Proceedings of the 1st International Workshop on Deep
Learning for mobile systems and applications, NY, USA, 2017
57. S. Yao, S. Hu, Y. Zhao, A. Zhang, T. Abdelzaher, DeepSense: a unified deep learning
framework for time-series mobile sensing data processing, in International World Wide Web
Conference Committee (IW3C2). Creative Commons CC BY, 2017
3 A Comprehensive Survey on Architecture for Big Data Processing. . . 49
58. A. Graves, Generating sequences with recurrent neural networks. ArXiv:1308.0850 (2013)
59. K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition.
arXiv:1512.03385 (2015)
60. J. Donahue, A. Hendricks, S. Guadarrama, M. Rohrbach, S. Venugopalan, K. Saenko, T. Darell,
Long-term recurrent convolutional networks for visual recognition and description, in CVPR,
2015
61. K. Greff, R.K. Srivastava, J. Koutnik, B.R. Steunebrink, J. Schmidhuber, LSTM: a search space
odyssey. arXiv:1503.04069 (2015)
62. N. Srivastava, R.R. Salakhutdinov, Multilodal learning with deep Boltzmann machines, in
NIPS, 2012
63. S. Bhattacharya, N.D. Lane, From smart to deep: Robust activity recognition on smartwatches
using deep learning, in PerCom Workshops, 2016
64. V. Radu, N.D. Lane, S. Bhattacharya, C. Mascolo, M.K. Marina, F. Kawsar, Towards
multimodal deep learning for activity recognition on mobile devices, in UbiComp: Adjunct,
2016
65. D. Figo, P.C. Diniz, D.R. Ferreira, J.M. Cardoso, Preprocessing techniques for context
recognition from accelerometer data. Pers. Ubiquit. Comput 14, 645–662 (2010)
66. C.-Y. Li, C.-H. Yen, K.C. Wang, C.-W. You, S.-Y. Lau, C.C.-H. Chen, P. Huang, H.-H. Chu,
Bioscope: an extensible bandage system for facilitating data collection in nursing assessments,
in UbiComp, 2014
67. E. Miluzzo, A. Varshavsky, S. Balakrishnan, R.R. Choudhury, Tapprints: your finger taps have
fingerprints, in MobiSys, 2012
68. W.D. Fisher, Machine learning for the automatic detection of anomalous events, PhD disserta-
tion, Colorado School of Mines, 2017, pp. 173
69. J. Laiho, M. Kylvaja, A. Hoglund, Utilization of advanced analysis methods in UMTS
networks, in IEEE Vehicular Technology Conference, Birmingham, USA, 2002, pp. 726730
70. J. Laiho, K. Raivio, P. Lehtimaki, K. Hatonen, O. Simula, Advanced analysis methods for 3G
cellular networks. IEEE Trans. Wirel. Commun. 4(3), 930942 (2005)
71. A.J. Hoglund, K. Hatonen, A.S. Sorvari, A computer host based user anomaly detection system
using the self-organizing map, in IEEE-INNS-ENNS International Joint Conference on Neural
Networks, vol. 5, Italy, 2005, pp. 411416
72. P. Lehtimaki, K. Raivio, A SOM based approach for visualization of GSM network perfor-
mance data, in International Symposium on Intelligent Data Analysis, Madrin, Spain, 2005
73. G. Chen, S.A. Jaradat, N. Banerjee, T.S. Tanaka, M.S. Ko, M.Q. Zhang, Evaluation and
comparison of clustering algorithms in analyzing ES cell gene expression data. Stat. Sin. 12(1),
241–262 (2002)
74. C.S. Hilas, P.A. Mastorocostas, An application of supervised and unsupervised learning
approaches to telecommunications fraud detection. Knowl. Based Syst. 21(7), 721–726 (2008)
75. J.H. Ward Jr., Hierarchical grouping to optimize an objective function. J. Am. Assoc. 58(301),
236–244 (1963)
76. J. You, Research of wireless network fault diagnosis based on Bayesian networks, in Second
International Symposium on Knowledge Acquisition and Modeling, vol. 3, 2009, pp. 59–64
77. I. Burguera, U. Zurutuza, S. Nadjm-Tehrani, Crowdroid: behavior-based malware detection
system for android, in ACM Workshop on Security and Privacy in Smartphones and Mobile
Devices, ACM, 2011, pp. 15–26
78. G. Dini, F. Martinelli, A. Saracino, D. Sgandurra, Madam: a multi-level anomaly detector for
android malware, in Computer Network Security, 2012, pp. 240–253
79. D. Arp, M. Spreitzenbarth, M. Hubner, H. Gascon, K. Rieck, C. Siemens, Drebin: effective and
explainable detection of android malware in your pocket, in NDSS, 2014
80. F. Al-Turjman, S. Alturjman, Confidential smart-sensing framework in the IoT era. J Super-
comput 74(10), 5187–5198 (2018)
81. M. Lindorfer, M. Neugschwandtner, C. Platzer, Marvin: efficient and comprehensive mobile
app classification through static and dynamic analysis, in 39th IEEE Annual Computer
Software and Applications Conference (COMPSAC), vol. 2, IEEE, 2015, pp. 422–433
Chapter 4
Taxonomy of Edge Computing:
Challenges, Opportunities, and Data
Reduction Methods
4.1 Introduction
In recent years, cloud computing has made significant advances and is being more
commonly used because of its high value efficiency and flexibility resulting from
consolidation. Cloud computing has led to the speedy development of cellular
internet and internet of things (IoT) packages for demanding situations using
sophisticated software and services. IoT is experiencing explosive growth and is
virtually transforming all aspects of modern life [1]. Cisco has predicted that 50
billion devices will be linked to the IoT by 2020 [2]. As these new technologies
enter the everyday lives of consumers, new challenges arise that cannot be safely
addressed through a centralized cloud computing structure, including stringent
latency, capacity constraints, useful resource-confined devices, uninterrupted ser-
vices with intermittent connectivity, improved security [3], and latency needs.
Because the records increase each day on the network, the performance of the
machines continues to be affected. This chapter discusses basic information about
cloud technology and IoT, associated challenges, and statistics reduction techniques
for edge computing to reduce data and latency time.
The term “cloud” refers to a network or internet that is present at a remote location.
The cloud can provide changes over public and private networks. Cloud comput-
ing is about controlling/moving, configuring, and using hardware and software
remotely. It offers online data storage and the basic functionality needed for a
business or society to operate. However, certain accommodations and models may
be needed locally for cloud computing to be feasible and accessible to end users.
Two categories of models for cloud computing are the deployment models and
service models.
The deployment models describe usage access to the cloud. The cloud may use
one of four types of deployment models: public, private, hybrid, or community:
• Public cloud: A public cloud sanctions systems and changes places to easily
accommodate the general public. A public cloud may be less secure than other
types of clouds because of its accessibility and openness.
• Private cloud: In a private cloud, systems and services are easy to access, use,
and understand within an organization. A private cloud is more secure than other
clouds because of its private nature.
• Hybrid cloud: A hybrid cloud has characteristics of both private and public
clouds. The analytical process is performed through a private cloud, whereas
non-analytical work is performed through a public cloud.
• Community cloud: A community cloud is deployed by a group of organizations
or industries.
The service models describe the eight types of services offered by a cloud:
• Everything-as-a-Service (EaaS): This type of service is given to all users of
the software and hardware to control business processes, including interactions
between users. The user only needs to have access to the internet.
• Infrastructure-as-a-Service (IaaS): In this type, users can avail a service or
computing platform by using a virtual platform that is connected to the network.
The service is automatically adjusted according to the use.
• Platform-as-a-Service (PaaS): In this service, the user is given a computing
platform along with operating software and other necessary programs.
• Software-as-a-Service (SaaS): Also known as “software on demand,” this service
deploys software on remote servers but also uses the internet.
• Hardware-as-a-Service (HaaS): This service uses uncovered hardware on which
a user can deploy its own software, similar to IaaS but deviating a bit.
• Workspace-as-a-Service (WaaS): This service is implemented by organizations
or companies that use cloud computing to set up the necessary workspace for
their employees.
• Data-as-a-Service (DaaS): This service is used to store large amounts of data
and information in a storage space.
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 53
IoT is an advanced automation and analytical system that has exploited various
aspects of technology, including recent advances in software enhancements, falling
hardware prices, and modern attitudes toward technology. IoT combines sensing,
communication, astronomically immense data, and artificial intelligence to dis-
tribute consummate systems for a product or accommodation. This collection of new
and advanced technologies has made major behavioral, political, and socioeconomic
changes in the lives of techno-savvy consumers. The most important feature of IoT
is its ability to make things “smart”; in this way, it enhances every aspect of life.
The “thing” in IoT can be any contrivance that is artificially smart enough
to communicate with a network without manual intervention to facilitate the
interaction. The communication can be with internal states of the “thing” or the
external environment, which in turn assists with the decision-making process. IoT
allows a “thing” to be smart enough to make a decision for you; for example,
your refrigerator and cabinets can become smart enough to detect when milk
and your favorite cereal are running low, then place an order with your preferred
grocer. As such, IoT introduces a new paradigm for active content, product, or
accommodation engagements. IoT devices have become smaller, more affordable,
and more powerful over time. IoT uses these purpose-built minuscule devices to
work with precision, scalability, and diversity.
The advantages of IoT span across every area of lifestyle and business to improve
the overall quality of life, including in communications, automation and control,
data gathering, data monitoring, time saving, cost-effectiveness, and efficiency.
Despite all its advantages, IoT also has security and privacy issues. The major
disadvantage of IoT is that it requires immense data transmission to the cloud and
thus increased security.
IoT is increasing exponentially and will potentially network billions or even
trillions of devices. Ericsson Inc. [4] has estimated that the data produced by
people, machines, and things will reach 500 zettabytes by 2019; however, Cisco
Global Cloud Index has estimated that global data centre IP traffic will only reach
10.4 zettabytes by that time [5]. Most IoT devices will be located at the edge of
the internet and could provide new applications, changing many aspects of both
traditional industrial productions and our everyday living. Some of these devices are
already on the market, including Apple watches, Oculus Rift helmets [6], Google
Nest [7], Fitbit sports trackers, and Google Glasses. By 2019, 45% of IoT-created
data will be stored, processed, analyzed, and acted upon close to (or at the edge of)
the network [8] but eventually transmitted to the cloud.
54 K. Jain and S. Mohapatra
As IoT grows, the data in the network is also increasing. To reduce these data,
the technology of edge computing may be used. Edge computing allows devices
to perform computations at the edge of the network, on downstream data on behalf
of cloud services and on upstream data on behalf of IoT services. Here, “edge” refers
to any computing and network resources that are between the data sources and cloud
data centers. For example, a smart phone is the edge between an individual and the
cloud, a gateway in a smart home is the edge between the house and the cloud, and a
micro data center and a cloudlet are the edge between a mobile device and the cloud.
Rationally, edge computing is performed at proximity to the data source. Figure 4.1
illustrates the concept of edge computing.
In the edge computing paradigm, the “things” are data consumers and data
producers. The edge can perform computing offloading, data storage, caching and
processing, request distributions, and distribution accommodations from the cloud
to the user. The edge has QoS requirements such as reliability, security, and privacy
firewalls.
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 55
IoT applications generate enormous amounts of data by IoT sensors. The big data
are subsequently analyzed to determine reactions to events or to extract analytics or
statistics. However, sending all of these data to the cloud would require prohibitively
high network bandwidth. Current research efforts are investigating how to better
exploit capabilities at the edge of the network to support the IoT and its needs [11].
In edge computing, the massive data generated by different kinds of IoT devices can
be processed at the network edge instead of transmitting the data to a centralized
cloud infrastructure due to bandwidth and energy consumption concerns. Edge
computing can provide services with faster response and better quality than cloud
computing. Edge computing is suitable for integration with IoT to provide efficient
and secure services for many end users. Edge computing-based architecture also
should be considered for future IoT infrastructure.
Push from Cloud Services Putting all computing tasks on the cloud is an efficient
way to process data because the cloud’s computing power outclasses the capability
of the things at the edge. However, compared with the fast-developing speed of
data processing, the bandwidth of the network has come to a standstill. With the
growing quantity of data generated at the edge, the speed of data transportation
is becoming a bottleneck in the cloud-based computing paradigm. For example,
about 5 gigabytes of data are generated by a Boeing 787 every second [12], but
the bandwidth between the airplane and either a satellite or the base station on the
ground is not large enough for data transmission. Consider an autonomous vehicle
as another example: 1 gigabyte of data is generated by the car every second, and
56 K. Jain and S. Mohapatra
it requires real-time processing for the vehicle to make correct decisions [13]. If
all the data needed to be sent to the cloud for processing, the response time would
be too long. Furthermore, it would be challenging for current network bandwidth
and reliability to support many vehicles in one area. In this case, the data needs to
be processed at the edge for shorter response times, more efficient processing, and
reduced network pressure.
Pull from IoT Many kinds of electrical devices will become part of the IoT. These
devices will play the roles of both data producers and data consumers, such as air
quality sensors, LED bars, streetlights, and internet-connected microwave ovens. It
is safe to infer that the number of things at the edge of the network will increase to
the billions in a few years. The raw data produced by these devices will be enormous,
with conventional cloud computing being too inefficient to handle the data. Thus,
most of the data produced by IoT will never be transmitted to the cloud. Instead, the
data will be consumed at the edge of the network.
In conventional cloud computing, data producers generate raw data and transfer
it to the cloud, and data consumers a send request for consuming data to the cloud.
However, this structure is not sufficent for IoT. First, the data quantity at the edge
is too large, which will lead to unnecessary bandwidth and computing resource
usage. Second, privacy protection requirements will pose an obstacle for cloud
computing in IoT. Lastly, most of the end nodes in IoT are energy-constrained
things, whereas the wireless communication module is usually very energy hungry.
Therefore, offloading some computing tasks to the edge could be more energy
efficient.
Change From Data Consumer to Producer In the cloud computing paradigm,
the end devices at the edge usually as as data consumers, such as when you
watch a YouTube video on your smart phone. However, people now are also
producing data from their mobile devices. This change from data consumer to data
producer/consumer requires more function placement at the edge, such as when
users take photos or record video.
This section discusses the basic techniques to implement edge computing. As edge
computing pushes the data storage and process near the edge of the network, small
data centers can be provided near the edge device, with or near the cloud, in
techniques known as cloudlets, mobile edge computing (MEC), and fog computing.
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 57
4.5.2.1 Cloudlet
and open MEC ecosystem. The last specification describes various metrics that can
potentially be improved by deploying a service on a MEC platform, such as latency,
energy efficiency, network throughput, system resource footprint, and quality [22].
Furthermore, the last specification also describes the best practices for measuring
such performance metrics.
ETSI announced six different MEC PoCs in September 2016, which were
accepted at the MEC World Congress in Munich. These PoCs will help to strengthen
the strategic planning and decision-making of organizations and identify which
MEC solutions may be viable in the network. Due to its advanced characteristics,
such as low latency, proximity, high bandwidth, and real-time insight into radio
network information and location awareness, MEC enables many new kinds of
applications and services for multiple sectors, such as consumer, enterprise, and
healthcare. MEC also seems to be a promising solution for handling video streaming
services in the context of smart cities [23].
The new technology of fog computing is similar to edge computing but it works
with the cloud. The OpenFog Consortium was founded to drive industry and
academic leadership in fog computing architecture, testbed development, and a
variety of interoperability and composability deliverables that seamlessly leverage
cloud and edge architectures to enable end-to-end IoT scenarios [24]. OpenFog
Consortium published a white paper on fog computing in February 2016, which
outlined its approach to an open fog computing architecture (OpenFog architecture)
[25]. The OpenFog Consortium defined fog computing as “a system-level hori-
zontal architecture that distributes resources and services of computing, storage,
control and networking anywhere along the continuum from Cloud to Things
(https://www.openfogconsortium.org).” Fog computing differs from edge comput-
ing by providing tools for distributing, orchestrating, managing, and securing
resources and services across networks and between devices that reside at the edge.
Edge architecture places servers, applications, and small clouds at the edge. Fog
computing jointly works with the cloud, whereas edge is defined by exclusion of
the cloud. Figure 4.4 illustrates the concept of fog computing. In Figs. 4.3 and 4.4,
fog computing and edge computing are shown as the middle layer; however, edge
computing is toward the edge device, whereas the middle layer works with the cloud
in fog computing.
Fog computing standardization is mainly administered by the OpenFog Con-
sortium, whose objective is to influence standard bodies to create standards so
that IoT systems at the edge can inter-operate securely with other edge and cloud
services in a friction-free environment. The OpenFog Consortium has set up six
working groups: an architecture working group, communications working group,
manageability working group, security working group, software infrastructure
working group, and testbed working group. These working groups evaluate, classify,
60 K. Jain and S. Mohapatra
and recommend standards, practices, and technologies that are appropriate for
OpenFog architecture to address corresponding challenges.
The expected huge increase in the number of IoT data sources, such as sensors,
embedded systems, and personal devices, provides good reason to implement
network-edge computing—that is, data pre-processing, local storage, and filtering
close to the data source. Specifically, data reduction at the network edge that is
on an IoT gateway device or a mini-server deployed locally (a cloudlet) or on
an IoT area network can prevent I/O bottlenecks, as well as dramatically reduce
storage, bandwidth, and energy costs. Many solutions have been provided, but these
solutions have two main obstacles in the implementation. Firstly, the most efficient
algorithms for data reduction in time series (which is one of the prevailing types of
data in IoT) were developed to work posteriori on big datasets; thus, they cannot
make decisions per incoming data item. Secondly, the state of the art lacks systems
that can apply any of the possible data reduction methods without adding significant
delays or heavyweight re-configurations.
There are many data reduction techniques used in edge computing. Based on
when data are sensed, two main categories of approaches can be used for IoT
data reduction at the network edge: (1) time series compression approaches or (2)
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 61
fusion center allocates network resources accordingly. By contrast, Jain and Chang
[30] adapted the sampling rate of the sensors to the communication resources while
minimizing the active sensors that stream data to a remote entity (i.e., server).
Compression Minimized datasets are easier to use in terms of processing and
inside the network movement for large storage. Compression-based techniques are
useful and appropriate for data reduction in edge computing. Compression can
be applied to already-sensed data. However, Li et al. [31] proposed a compressed
sensing framework, in which compressed sensing is applied prior to data acquisition.
In an approach by Liu et al. [32], compressed sensing is applied to the data on-the-
fly, under the assumption of opportunistic routing.
Data Aggregation Data aggregation consists of considering diverse sources of data
and gathering them together to build an accurate representation of the phenomena
under observation. These sources differ in that each of them is responsible for
a specific sensing task. In IoT and wireless sensor networks, data aggregation
techniques include operations such as solving the maximum, minimum, sum, aver-
age, median, and the count. Aggregation methods can be divided into centralized
and distributed methods. Centralized methods require continuous communication
among network entities. Hence, the communication overhead incurs additional
costs; consequently, such methods are not suitable for sensor networks. Therefore,
distributed methods such as clustering, multipath, and aggregate trees are commonly
employed [33]. Bash et al. [34] performed data aggregation with the purpose of
minimizing messaging costs among the sensor nodes. The node from which the
data is aggregated is proportional to the given size of the network. Similarly, Lin et
al. [35] divided the sensor network into non-overlapping regions, then sampled and
aggregated data from each region. Moreover, Kimura and Latifi [36] summarized
several methods of data compression, which consists of transmitting data as little
as possible by aggregating. Such methods are often valid for specific lookup
applications; hence, they are not applicable for applications that rely on fast- and
ever-changing sensory data.
Most data aggregation methods are based on trees or other fixed data structures.
For example, Huang et al. [37] proposed a spanning tree-based data aggregation
mechanism for large-scale networks. Each leaf in the tree is responsible for sensing
data from a given location. Data aggregation starts from the leaves and propagates
to the root (i.e., data collection entity). Other data aggregation methods rely on
no specific data structures [38]. Multi-task data sharing on the network leads to
efficient use of the available bandwidth; however, it introduces more computation
and communication costs. Such a technique aims at collecting as little data as
possible to satisfy all tasks’ needs.
IoT network architectures usually have a sensor-gateway router-cloud structure
[39]. Indeed, such an architecture tailors the proposed mechanism for data reduction,
with an aim to reduce the amount of transmitted data (and hence transmission
power) at a lower time cost. Pratt and Fink [40] proposed a technique for fast
compression of time series data and indexing of the compressed series. The
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 63
compression mechanism identifies the important points of a series and discards the
remaining ones. This technique has a good performance and can also be applied to
resource-concerned applications. This solution proposed a framework for real-time
data filtering over two tiers—namely, the gateway and the edge tier.
Other Approaches Other methods are being used to optimize data redundancy and
duplicity. Some techniques use the topological structures of unorganized data and
minimize the overall data using the network theory approaches:
1. Network Theory: The graph or networks play a basic role in reduction techniques
of higher dimensional unorganized data into lower dimensional organized data
[41]. However, the extraction of these structures is a difficult task due to
heterogeneity and complex structures. Trovati [42] proposed a network theory
approach that can abstract the properties of a network from the data. The
topological networks are constructed by establishing and evaluating relationships
(links) among different data points. The statistical node analysis of the networks
is performed for optimization and immensely colossal data reduction [43]. The
optimized networks are used in small networks, free scale, and any random
network.
2. Data Redundancy Elimination: Data redundancy is a key issue for data anal-
ysis in immensely colossal data environments. Three main reasons for data
redundancy are the integration of nodes, the expansion of datasets, and data
replication. The integration of a single virtual machine (VM) brings around
97% more redundancy, and the magnification in immensely colossal datasets
comes with 47% redundant data points [44]. Cluster deduplication may be
a generalized massive information reduction scheme for disk-based cluster
backup systems. The redundant information held on multiple disks and partition
areas is a serious challenge for giant processing systems. The deduplication
techniques salow systems to handle completely different information chunks
(partitions) and exploitation hash functions with lower intra-node and inter-node
communication overheads. Additionally, these strategies improve the storage
potency by eliminating redundant information from multiple nodes.
3. Data Pre-processing: Data pre-processing techniques help to extract meta-data
for further processing. Di Martino et al. [45] discussed the basic approaches for
data pre-processing, which are generally dependent on semantic analysis or data
structures. Some other techniques are also present, such as low memory pre-
filters for streaming data, URL filtration, and map-reducing methods.
4. Data Mining and Machine Learning: Recently, many data mining and machine
learning techniques have been used for data reduction. This concept is either
implemented to reduce data instantly after the acquisition or customized to
address some specific tasks. Jiang et al. [46] proposed a context-aware big data
forwarding scheme using distributed wearable sensor devices. The proposed
technique minimizes the communication and storage overhead in big data.
64 K. Jain and S. Mohapatra
Papageorgiou et al. [47] presented a new data reduction solution that automates the
switching between different data handling algorithms at the network edge, including
an analysis of adjusted data reduction methods. The authors also proposed three
versions of a new algorithm that is capable of performing real-time reduction of
incoming time series items based on the concept of perceptually important points.
Rehman et al. [48] presented and evaluated a novel big data processing archi-
tecture named RedEdge (i.e., data reduction on the edge) that incorporates a
mechanism to facilitate the processing of big data streams near the source of the
data. The RedEdge model leverages mobile IoT-termed mobile edge devices as
primary data processing platforms. In the case of the unavailability of computational
and battery power resources, it offloads data streams in nearer mobile edge devices
or to the cloud. The RedEdge architecture and the related mechanism were evaluated
within a real-world experiment setting involving 12 mobile users.
Kim et al. [49] took the approach of moving control of a low-power wide
area network (LPWAN) to the edge cloud, which provides computing and storage
environments at base stations. An LPWAN gateway is then integrated with the edge
cloud, and LPWAN data are cached at the edge cloud in the gateway for LPWAN
control. LPWAN controls, such as report interval control for data, transmission
power control, and data aggregation, are applied to improve the efficiency of data
transmission. Network control is performed by using cached data in the edge cloud.
Compared with the existing LPWAN control approach, the proposed approach
exhibits improved performance for IoT data transmission. The simulation results
demonstrated the proposed approach’s efficiency.
Naming The naming system is very important for programming, addressing, and
identification, among others. A naming scheme is required to handle the mobility
of things, dynamic network topology, and security issues in edge computing.
Traditional naming systems, such as DNS, are not able to adequately handle
dynamic edge computing. To overcome this issue, new naming mechanisms, such
as named data network (NDN) [55] and MobilityFirst [56], can be deployed in edge
computing. NDN provides a hierarchical structured name for the data network and
is user friendly. MobilityFirst can separate the name from the network address to
provide better mobility support and is efficient when applied to the edge service.
Data Abstraction With the use of IoT, a huge number of data generators are
present in the data network. Human interaction in edge computing should be
reduced; the edge links and nodes should use all the data and interact with the users.
In this case, the data should be pre-processed at the gateway. Collecting the data
will serve the application, and the process should be allowed to control the things
to complete the various services. Combining data representation and operations as
a layer of data abstraction serves as a public interface for all things connected with
the edge operating system.
66 K. Jain and S. Mohapatra
4.10 Conclusion
Today, many devices have been pushed from the cloud to the edge of the network.
When data are processed at the edge, shorter response times and greater reliability
can result. Bandwidth also can be saved when a larger part of data is handled at the
edge rather than uploaded to the cloud. The IoT and mobile devices have driven the
trend of edge computing techniques from data consumer to data producer/consumer,
where it is efficient to compute the data at the edge of the network. This chapter
discussed the basics of edge computing, its various methodologies, and some
data reduction techniques. Finally, we presented some of the challenges of edge
computing and solutions to overcome them, including programmability, naming,
data abstraction, privacy, and security.
References
3. M. Chiang, T. Zhang, Fog and IoT: an overview of research opportunities. IEEE Internet Things
J. 3, 854–864 (2016)
4. Ericsson Inc, Ceo to shareholders: 50 billion connections 2020, p. 1, (2010). Available at: http:/
/www.ericsson.com/thecompany/press/releases/2010/04/1403231
5. Cisco global cloud index: Forecast and methodology, 2014–2019 white paper. (2014)
6. Oculus, Oculus Rift helmet: next generation virtual reality. (2016). Available at: https://
www3.oculus.com/en-us/rift/. 2
7. Google, Nest IoT devices. (2016). Available at: https://nest.com/.3
8. D. Evans, The Internet of Things: how the next evolution of the Internet is changing everything.
CISCO White Paper, vol. 1, pp. 1–11, 2011.10
9. F. Wortmann, K. Flüchter, Internet of Things. Bus. Inf. Syst. Eng. 57(3), 221–224 (2015)
10. https://vtechworks.lib.vt.edu/bitstream/handle/10919/78767/Kalin_JH_T_2017.pdf?sequence=1
11. A. Brogi, S. Forti, QoS-aware deployment of IoT applications through the fog. IEEE Internet
Things J. 4, 1185–1192 (2017)
12. Boeing 787s to Create Half a Terabyte of data Per Flight, Says Virgin Atlantic. Accessed on 7
Dec 2016 [Online], Available : https://datafloq.com/read/self-driving-cars-create-2-petabytes-
data-annually/172
13. Self-Driving Cars Will Create 2Petabytes of Data, What are the Big\Data Opportunities
for the Car Industry? Accessed on 7 Dec 2016. [Online]. Available: http://
www.computerworlduk.com/news/data/boeing-787s-create-half-terabyte-of-data-per-flight
says-virgin-atlantic-3433595/
14. M. Satyanarayanan, P. Bahl, R. Caceres, N. Davies, The case for VM-base cloudlets in mobile
computing. Pervasive Comput. 8, 14–23 (2009)
15. ETSI, Mobile-edge Computing Introductory Technical White Paper, White Paper, Mobile-
edge Computing Industry Initiative, (2014). https://portal.etsi.org/portals/0/tbpages/mec/docs/
mobile-edge_computing_–_introductory_technical_white_paper_v1
16. Y.C. Hu, M. Patel, D. Sabella, N. Sprecher, V. Young, Mobile Edge Comput-
ing a Key Technology towards 5G, (2015). http://10.3.200.202/cache/8/03/etsi.org/
6e14a9668574b8b935111768d9f6e501/etsi_wp11_mec_a_key_technology_towards_5g.pdf
17. ETSI GS MEC 001, Mobile Edge Computing (MEC) Terminology V1.1.1, (2016). http://
www.etsi.org/deliver/etsi_gs/MEC/001_099/001/01.01.01_60/gs_MEC001v010101p.pdf
18. ETSI GS MEC 002, Mobile Edge Computing (MEC) Technical Requirements
V1.1.1, (2016). http://www.etsi.org/deliver/etsi_gs/MEC/001_099/002/01.01.01_60/
gs_MEC002v010101p.pdf
19. ETSI GS MEC 003, Mobile Edge Computing (MEC) Framework and Reference Archi-
tecture V1.1.1, (2016). http://www.etsi.org/deliver/etsi_gs/MEC/001_099/003/01.01.01_60/
gs_MEC003v010101p.pdf
20. ETSI GS MEC-IEG 004, Mobile Edge Computing (MEC) Service Scenarios
V1.1.1, (2015). http://www.etsi.org/deliver/etsi_gs/MEC/001_099/004/01.01.01_60/
gs_MEC003v010101p.pdf
21. ETSI GS MEC-IEG 005, Mobile Edge Computing (MEC) Proof of Concept Frame-
workV1.1.1,(2015). http://www.etsi.org/deliver/etsi_gs/MECIEG/001_099/005/01.01.01_60/
gs_MEC-IEG005v010101p.pdf
22. ETSI GS MEC-IEG 006: Mobile Edge Computing Market Acceleration MEC Metrics Best
Practice and Guidelines V1.1.1, (2017) http://www.etsi.org/deliver/etsi_gs/MECIEG/001_099/
006/01.01.01_60/gs_MEC-IEG006v010101p.pdf
23. T. Taleb, S. Dutta, A. Ksentini, M. Iqbal, H. Flinck, Mobile edge computing potential in making
cities smarter. IEEE Commun. Mag. 55, 38–43 (2017)
24. Open Fog Consortium, [Online]. https://www.openfogconsortium.org/
25. OpenFog Consortium Architecture Working Group, OpenFog Architecture Overview
White Paper, https://www.openfogconsortium.org/wp-content/uploads/OpenFog-Architecture-
Overview-WP-2-2016.pdf
68 K. Jain and S. Mohapatra
26. Q. Shen, L. Huang, G. Zhang, J. Gong, Policy Control and Traffic Aggregation for M2M
Services in Mobile Networks. in International Conference on Mechatronic Sciences, Electric
Engineering, and Computer (MEC ’13), pp. 3391–3395 (2013)
27. Cisco. Krikkit open source software, Feb 2014. http://eclipse.org/proposals/technology.krikkit/
28. H. Zou, Y. Yu, W. Tang, H.-W. M. Chen. FlexAnalytics: a flexible data analytics framework
for big data applications with IO performance improvement. Big Data Res. J. 1, 4–13 (2014).
Special Issue on Scalable Computing for Big Data
29. R. Willett, A. Martin, R. Nowak. Backcasting: adaptive sampling for sensor networks. in
Proceedings of the 3rd international symposium on Information processing in sensor networks.
ACM, pp. 124–133. (2004)
30. A. Jain, E.Y. Chang, Adaptive sampling for sensor networks. in Proceeedings of the 1st
international workshop on Data management for sensor networks: in conjunction with VLDB
2004. ACM, pp. 10–16, (2004)
31. S. Li, L. Da Xu, X. Wang, Compressed sensing signal and data acquisition in wireless sensor
networks and internet of things. IEEE Trans Ind Inf 9(4), 2177–2186 (2013)
32. X.-Y. Liu, Y. Zhu, L. Kong, Y.G. Cong Liu, A.V. Vasilakos, M.-Y. Wu, CDC: Compressive data
collection for wireless sensor networks. IEEE Trans Parallel Distributed Syst 26(8), 2188–2197
(2015)
33. F. Al-Turjman, Fog-based caching in software-defined information-centric networks. Comput
Electr Eng J 69(1), 54–67 (2018)
34. B.A. Bash, J.W. Byers, J. Considine. Approximately uniform random sampling in sensor
networks. in Proceeedings of the 1st international workshop on Data management for sensor
networks: in conjunction with VLDB 2004. ACM, pp. 32–39, (2004)
35. S. Lin, B. Arai, D. Gunopulos, G. Das, Region sampling: Continuous Adaptive Sampling on
Sensor Networks, IEEE 24th International Conference on Data Engineering, (2008), https://
ieeexplore.ieee.org/document/4497488/
36. N. Kimura, S. Latifi, A survey on data compression in wireless sensor networks. in Interna-
tional Conference on Information Technology: Coding and Computing (ITCC’05)-Volume II,
Vol. 2. IEEE, 8–13, (2005)
37. Z. Huang, W. Lu, K. Yi, Y. Liu, Sampling based algorithms for quantile computation
in sensor networks. in Proceedings of the 2011 international conference on Manage-
ment of data – SIGMOD ‘11. ACM Press, New York, New York, USA, 745. (2011).
https://doi.org/10.1145/1989323.1989401
38. K.-W. Fan, S. Liu, P. Sinha, Structure-free data aggregation in sensor networks. IEEE Trans.
Mob. Comput. 6(8), 929–942 (2007)
39. A. Papageorgiou, B. Cheng, E. Kovacs, Real-time data reduction at the network
edge of Internet-of-Things systems. in 2015 11th International Conference
on Network and Service Management (CNSM). IEEE, pp. 284–291. (2015).
https://doi.org/10.1109/CNSM.2015.7367373
40. K.B. Pratt, E. Fink, Search for patterns in compressed time series. Int. J. Image Graph. 2,
89–106 (2002)
41. J.W. Patty, E.M. Penn, Analyzing big data: social choice and measurement. Polit. Sci. Polit.
48(01), 95–101 (2015)
42. M. Trovati, Reduced topologically real-world networks: a big-data approach. Int. J. Distrib.
Syst. Technol. (IJDST) 6(2), 13–27 (2015)
43. M. Trovati, N. Bessis, An influence assessment method based on co-occurrence for topologi-
cally reduced big data sets. in Soft Computing, pp. 1–10, (2015)
44. B. Jalali, M.H. Asghari, The anamorphic stretch transform: putting the squeeze on “big data”.
Opt. Photonics News 25(2), 24–31 (2014)
45. B. Di Martino et al., Big data (lost) in the cloud. Int. J. Big Data Intell. 1(1–2), 3–17 (2014)
46. P. Jiang et al., An intelligent information forwarder for healthcare big data systems with
distributed wearable sensors. IEEE Syst. J. 99, 1–9 (2014)
47. 2015 11th International Conference on Network and Service Management (CNSM) in
[Real-Time Data Reduction at the Network Edge of Internet-of-Things Systems Apostolos
4 Taxonomy of Edge Computing: Challenges, Opportunities, and Data. . . 69
Papageorgiou, Bin Cheng, Erno Kovacs ¨ NEC Laboratories Europe Heidelberg, Germany
[email protected], [email protected], [email protected]]
48. M.H. Rehman, P.P. Jayaraman, S.R. Malik, A.R. Khan, M.M. Gaber, RedEdge: A Novel
Architecture for BigData Processing in Mobile Edge Computing Environments. J. Sens.
Actuator Netw. 6(3), 17 (2017). https://doi.org/10.3390/jsan6030017
49. D.Y. Kim, S. Kim, J.H. Park, A combined network control approach for the edge cloud and
LPWAN-based IoT services. (2017). https://doi.org/10.1002/cpe.4406
50. C. Yang et al., A spatiotemporal compression based approach for efficient big data processing
on Cloud. J. Comput. Syst. Sci. 80(8), 1563–1583 (2014)
51. K. Ackermann, S.D. Angus, A resource efficient big data analysis method for the social
sciences: the case of global IP activity. Procedia Comput. Sci. 29, 2360–2369 (2014)
52. B.H. Brinkmann et al., Large-scale electrophysiology acquisition, compression, encryption,
and storage of big data. J. Neurosci. Methods 180(1), 185–192 (2009)
53. M. Weinstein et al., Analyzing big data with dynamic quantum clustering. arXiv preprint
arXiv:1310.2700, (2013)
54. A. Cichocki, Era of big data processing: a new approach via tensor networks and tensor
decompositions. arXiv preprint arXiv:1403.2048, (2014)
55. L. Zhang et al., Named data networking (NDN) project. Xerox Palo Alto Res. Center, Palo
Alto, CA, USA, Tech. Rep. NDN-0001, (2010)
56. D. Raychaudhuri, K. Nagaraja, A. Venkataramani, MobilityFirst: a robust and trustworthy
mobility-centric architecture for the future Internet. ACM SIGMOBILE Mobile Comput.
Commun. Rev. 16(3), 2–13 (2012)
57. F. DaCosta, Rethinking the Internet of Things: A Scalable Approach to Connecting Everything
(ApressOpen, New York, NY, 2013)
Chapter 5
Applications of the Internet of Things
with the Cloud Computing Technologies:
A Review
The term “Internet of Things” came into existence from 1999. Internet of Things
has the power to change the world and is a system of interconnection between
uniquely identifiable computing devices and mechanical and digital machines. This
interconnection is made without involving human-to-human or human-to-machine
interaction, and instead the data are transferred over the network. Internet of Things
is a general concept where the devices could gather the data around the world
and could share the data over the network where it could be utilized for various
purposes. Wikipedia states, “The Internet of Things (IoT) is the interconnection
of uniquely identifiable embedded computing devices within the existing Internet
infrastructure”; i.e., it connects the embedded system to the internet. In recent
years, there is an enormous development in the IoT field. Embedded systems are
much attractive because of their low cost, low space, low power consumption, and
autonomous factors. It also includes other factors such as I/O speed and cost, energy
consumption, etc., based on the instructions.
Figure 5.1, obtained from Wikipedia, depicts the technological roadmap for
Internet of Things. From this figure, it can be observed that by 2000, RFID tags
were used to track the routing of the inventory to prevent the loss of logistics. This
paved way for increase in the demand for logistics by providing support for the
supply chain helpers. By the mid of 2000–2010, vertical-market applications such
as surveillance, security, healthcare, transport, food safety, document mapping have
advanced sensor
Miniaturisation, power- fusion
efficient electronics, and
available spectrum
Teleoperation and
telepresence: Ability to
monitor and control
Ability of devices located distant objects Physical-World
indoors to receive Web
geological signals
Locating people and
Cost reduction leading
everyday objects
to diffusion into 2nd Ubiquitous Positioning
wave of applications
Surveillance, security,
healthcare, transport,
Demand for expedited food safety, document
logistics management
Vertical-Market Applications
RFID tags for
tacilitating routing, Source: SRI Consulting Business Intelligence
inventorying, and loss
prevention Supply-Chain Helpers
Fog and edge computing look similar in terms of intelligence and also in
the way the data are processed. But they differ in identifying the location of
the intelligence and also in power computation. In edge computing environment,
intelligence is placed in devices like embedded automation controllers, whereas
fog computing environment places intelligence in the local area network (LAN)
itself. For processing the power, fog computing uses edge devices and the gateways
along with the LAN, but in the case of edge, the processing power is placed with
the intelligence. Hence, the most challenging features in this computing are power
consumption, latency, security, and also the data size which can be distributed across
the network.
Older IoT technologies include a virtually infinite number of endpoints con-
nected to the network. Because of this advantage, data collection and processing
has become a biggest challenge. In order to solve this challenge, edge computing
is introduced. The computation processes which are performed at the nodes have
given a feel as if it is performed at the data source itself. Edge computing can also
be stated as an extension of older technologies. The extended features available
in the edge computing are peer-to-peer networking, distributed data, self-healing
network technology, and remote cloud services. Improved hardware security and
also lowering the power requirement are satisfying factors that are provided by the
processors of edge computing systems. This edge computing system also provides a
highly optimal solution as the hardware is embedded with the flash-storage arrays.
The concept of interaction of the centralized system with embedded systems and
74 U. Ram Jagannath et al.
Table 5.1 gives a short description of the IoT applications which will be discussed
in detail.
IIoT describes the usage of Internet of Things across several industrial activities like
manufacturing, logistics, metals, and mining and in other sectors. Internet of Things
76 U. Ram Jagannath et al.
IoT brings a number of applications for retailers for improving store operations,
minimizing theft occurrences; increasing customer purchase by selling more num-
ber of the products to them, and also producing a new inventory product [7]. It will
make the customer know everything in the store and aid them closer to the buying
decision. RFID tags are attached to every product in the store, and its availability
is checked periodically. Customers gain the information about the goods through
Bluetooth beacons or other digital receivers. The digital signature will navigate the
customer to the product. The following actions can be performed for improvising
the quality of purchase and make the customer think and take an easy decision
in buying a product. From the analytics, the retailer will know about customer’s
perspectives which will later help in improvising their product according to the
customer’s interest.
• Interactive digital signature screens
• Mobile shopping applications
• Sensor-based items tracking
• Tailored digital marketing
• Optimized inventory management
5 Applications of the Internet of Things with the Cloud Computing. . . 77
In [15], the authors have proposed a cloud-centric IoT architecture for supply
chain management (SCM). In traditional SCM, the movement of physical objects
was performed directly, whereas, in this paper, the authors have controlled and
coordinated the supply chain process based on object sensing. For implementing
this framework, they have used proposed layers in their framework, namely, object
management, cloud computing platform, information as a service, and smartphone
user. Thirty-one percent of manufacturers are implementing Internet of Things
enhancements to their internal operations, with another 56% exploring doing so
in order to cut operational costs, achieve supply chain efficiencies, and improve
predictive maintenance capabilities. Operations performed with these layers include
submission of request by the user to the cloud; the cloud computing platform will
make response to the user by identifying the location; this cloud computing platform
helps the user to update the data in the database and hence the user can publish the
sensed information; with the received information, the cloud will perform operation
and will return the user about the task which is mapped according to the physical
location of the user; and finally all the transactions and the conversations are stored
to the cloud for future references.
Internet of Things paves a great and fresh way for the car manufacturers to upgrade
their cars in a highly competitive market. As stated in [2] by Kevin Ashton, this
also enhances the idea, where a person can drive around the world and can stay
connected all the time. Before the invention of IoT in automobiles, one has to search
for the route manually, while the IoT has made it more convenient. For example, a
person could set the destination place using Ground Positioning Systems (GPS),
and the driver will be directed to the destination place and this also needs some
improvements.
Auto-pilot mode is the most modern technology introduced in recent times. High-
resolution cameras can be used to detect soft and hard objects in the lane and also in
the vehicle surroundings. Forward-facing radar with enhanced processing collects
additional data about the world on a redundant wavelength that helps the user to
see through heavy rain, fog, dust, and even the car behind and ahead. This could
also aid in detecting the traffic signals and signs that are already inbuilt in the
system and could drive the car as safe as possible. It can also adapt to the changes
in the environmental roads around the world, and it could also offer lane change
automatically. IHS Markit predicts that more than 70 million connected cars will be
on the road by 2023.
Designing a connected vehicle platform on cloud IoT core will be a better
solution for managing the connected cars by making a use of cloud IoT core on
Google Cloud Platform (GCP). The use cases that can be defined by making a
combination with data storage and platforms are usage-based insurance, predictive
maintenance, freight tracking, and customized in-vehicle experience. The main
challenges that a development team will face in this domain are device management,
data ingestion, data analytics, applications, and predictive models. Besides these
challenges, the developer must also focus on security and set the boundary level for
the applications they design that are implemented in the cloud-based system.
Self-driving cars are learned from the data cloud, the collections of data are brakes,
tire and transmission performance, fuel consumption, and efficiency of the vehicle.
The main reason for self-driving is high safety, and it will abide all traffic rules
and regulation as shown in [5]. Averagely, parking coverage takes 31% of land use
in big cities, like San Francisco, and even more, 81% in Los Angeles and 76%
in Melbourne, while at the lower end we find New York (18%), London (16%),
and Tokyo (7%). Hence, the removal of unnecessary parking areas is the first task
for municipalities to create a better urban planning. The data are provided with
high security, and the hackers cannot access the data easily without the user’s
authorization. There exists confusion when the user tries parking his or her vehicle,
80 U. Ram Jagannath et al.
and it leads to great pressure and stress to the user which also results in waste of
time. As self-driving vehicle is automatically parked in the parking area without any
confusion it reduces pressure as well as the time for user. IoT predicts the current
scenario and performs the action according to the problem.
In [16], the authors have designed a smart parking system. In this system, the
authors have specified that the system can be implemented for any type of parking
like parking available at covered parks, open parks, and also street side parking. For
this system, the authors have designed an architecture which includes cloud service
provider. The main function of this cloud service provider is to collect information
about the parking area and store the information in the cloud. The major components
of this smart parking system are centralized server, Raspberry Pi, image capturing
device, a navigation system, displaying device, and the user device. The centralized
server acts as a database that will hold the entire information about the parking
system, some of the information includes a number of parking slots, availability of
vehicles, and so on. Raspberry Pi is a microcontroller used to implement the parking
system with a camera attached to it. This Pi-camera is used to capture pictures of the
parking area for validating the area to check the availability of slots in that area. The
main role of the navigation system is to navigate the users to the available parking
lot which is nearer to his or her current location. Displaying device is a monitor or
tab available at the admin’s location to monitor the device and to modify the parking
lot if needed. The user device is used to connect the user with the parking system
using their smartphones or browsers.
Safety of life is more important when compared to any other application. Records
show that many lives were exploited by fire accidents, and also it takes a longer
duration for the fire service to reach the destination. To avoid these types of
accidents, many companies are using IoT devices for safety. It is safe in many ways
as it is more evident that if fire accident occurs, it automatically connects itself to
the hot-connect or cloud environment and it directly sends a voice or text message
to the concerned fire agency. The working principle is not like the normal smoke
detectors which could only give beeping alarm sounds but it also helps to save
people. This has battery power that is estimated to last up to 10 years and also it
is not a big bottleneck to change the battery. It is more rapidly replacing the normal
smoke detectors as this proves to be more useful and also available at an affordable
cost. One thousand smoke detector surveys were mailed, covering about 70 percent
of the town’s households. While the results staggered in over a period of 3 months,
315 completed surveys (31.5 percent) were ultimately returned. The high return rate
can be interpreted as a strong interest on the part of the public concerning fire safety.
In the industrial sector, the smoke detectors play a different role, where it is used
as the contamination detector. The detector is placed at the chimney and monitors
the smoke that is being passed out through it. If the contaminants are greater than
5 Applications of the Internet of Things with the Cloud Computing. . . 81
the recommended level, it will send a warning message to the main supervisor and
also to the concerned person so that it will never be ignored. It ensures green living
to all the people and greatly reduces pollution.
As described by Gauer, A., Smart City Architecture and its applications based on
IoT in [4], smart cities improve the quality of lifestyle by improving infrastructures
like delivering clean water, dependable power, and efficient public lighting. The
main motive of a smart city is to provide more efficient water supply, an innovative
solution to traffic congestion, and more reliable public transportation. Water can be
highly saved by providing the following facilities.
• Leakage detection in pipes
• Wastewater recycling
• Managing the storm-water
The benefits of the smart city include increased safety, reduced traffic, lower
levels of pollution, more efficient use of energy, and improve the overall quality of
life for future city dwellers. The main aim of the smart cities is extended to keep the
citizens safe, and it should provide education and technological job opportunities to
the citizens with open innovation. Gujarat International Finance Tec-City stands as
a great example for smart cities. It has better city planning and development, and
internet facility is provided all over the city with a faster supply of products at a
lower operating expense. The major issue stands in the contamination of air, and to
resolve this issue many green buildings are set up.
The companies which focus on security management can be connected with
smoke detectors so that they can protect the buildings from fire. The device will
send the message to the cloud-based dashboard only if the fire is detected or if
the system is in need of maintenance. This will greatly reduce the man-hours in
performing the maintenance check frequently, and even the process may not be
needed. These connected smoke detectors are more advantageous when compared
to the traditional system. These smoke detectors can be connected to devices such
as desktops, laptops, tablets or smart-phones to get alert messages which proves
helpful for fire security personnel.
5.2.10 Construction
attached to the devices so that tracking can be made easy. The cost of the equipment
is comparably higher than earlier devices. This will help the consumer to buy the
equipment only once in a lifetime, instead of purchasing frequently whenever it is
missed. Repair and service of the equipment can be monitored, and the problem
could be resolved with a warning alert from the equipment. Energy consumption
by the equipment can be minimized by the sensors which will adjust the actions
to be performed by the tools automatically. The sensors will automatically switch
on/off to save energy. The main advantage is the quality work is increased and
more convenient for users. The construction project work can be done on time,
and the quality of the construction can be improved. The future Internet of Things
will provide intelligent building management systems which can be considered as
a part of a much larger information system used by facilities managers in buildings
to manage energy use and energy procurement and to maintain efficient building
management systems.
A smart panel can perform multiple functions like tracking the electrical usage and
wastage in a detailed manner, and it also displays the result on the computer screen.
They are simple, flexible, and have an open architecture which proves that they
5 Applications of the Internet of Things with the Cloud Computing. . . 83
are simple in construction and implementation. Using these applications, the user
can remotely access the electrical units which are installed far away which help in
reducing the electrical energy.
Another type is smart solar panels, and these have multiple abilities which could
perform different actions depending on the conditions required. For example, this
panel can be used to convert solar energy into electrical energy or to the required
form of energy in which it acts as a transducer in one form and it could convert air
into drinking water from the sunlight and air in another form. All the installed panels
are monitored by the organization. The purity and taste of the water are obtained
by adding the required minerals to it. It is connected and so it provides optimized
performance. It could provide an average of 8–20 bottles of water per day. The
main advantage of using these panels is they need not be hooked up to the home’s
electrical systems, and they use full renewable resources to gain energy.
Energy is a very important aspect for any household, industries, agriculture, and
so on. Managing energy source and conserving it for better performance is a
challenging task. IoT paradigm promises to increase the visibility and awareness
of energy consumption, thanks to smart sensors and smart meters at the machine
and production line level. Since every appliance depends on energy supply, energy
plays a vital role in day-to-day life. An interface like Arduino Microcontrollers
controls the usage of an appliance like the intensity of light and the speed of
the fan. Humidity, temperature, and light intensity are taken into consideration in
developing a better IoT system for energy management. IoT devices are controlled
and connected for small cities and communities, transportation systems, buildings,
lighting systems, factories, and more devices. Building Energy Management Sys-
tems (BEMs) attracted $1.4 B in VC Funding from 2000 to 2014 (26% of all
investment in building energy technology). In 2020, about 77% of the $2.14 billion
US market will comprise BEMS applications, and 40% will come from buildings
below 50,000 square feet. The US market for sensors and controls for BEMs will
rise at a 17% compound annual growth rate to $2.14 billion in 2020. Possible future
directions for energy management are as follows:
• Energy-efficient mechanisms for software-defined IoT solutions, which can
provide scalable and context-aware data and services.
• Fog computing can lead to energy saving for most of the IoT applications;
therefore, it is important to study energy consumption of fog devices for IoT
applications.
84 U. Ram Jagannath et al.
Wearable devices are considered as the hot topic these days. A reduced job is much
preferred in every field which resulted in this technology. Some of the important
functions to be considered for a smart wearable in the industrial location is it could
sense cut-off radiation level and will alert the user when the environment becomes
worse.
Health- and fitness-related devices are also available these days which could
monitor the heartbeat rate, blood pressure, footsteps count, sleep monitor, etc. Some
of the useful features are it could find the body temperature by which the user could
get an early indication of cold or flu. Most of the users have a question why to
use a wearable gadget while a single smartphone could do all these work. It is
not comfortable to take our smartphone for every transaction, but the smartphone
will act as the central unit of the wearable by connecting it through Bluetooth or
the network which grants access for the wearable to surf the data available in the
smartphone.
Further, wearable reduces manual work to a great extent where the devices in
the home are being connected and the user can control any of the devices which
are connected in the same network without requiring any movement from place to
place.
Biosensors are gaining more interest nowadays, and they promise to be one of
the constant growing fields in IoT [9]. Biosensors are mainly used in the field
of healthcare, applications related to sports, and mostly in the military. The main
advantage of this application is low cost and also provides real-time information.
The term “wearable biosensors” (WBS) is evolved by combining both the smart
wearable and the biosensors in a single unit. The required sensors are installed in
the various parts of the body so they could be monitored all the time, without any
whatsoever disturbances. Their size will be very small which makes them easy to
carry wherever the user moves. Based on the usage of such sensors, these devices
are popularly named as immunosensors, glucometers, biocomputers, etc., some of
the examples include (1) Google lens (2) ring sensor and (3) smart shirt.
In [13], the authors have proposed a device named Cellcon. This device is
developed based on the image analysis concept which is portable and also affordable
to do the process. This is a hardware device with a ball lens and a mobile application.
The authors designed this application to do the cell count and to identify the
presence of the malarial parasite. The blood sample is collected in a slide and a photo
is taken by a ball lens that is attached to the phone clip, and hence a magnified image
of the blood sample is obtained. This IoT device uses cloud storage for storing the
image data, retrieve the data, and perform the analysis through and from the internet.
The image is compared with the images in the database and the results are sent to
the patient. In this application, the first half of the application, i.e., collecting the
blood sample and taking a photo, is performed at the patient’s location. The patient
may reside at any remote location, and the image data will be passed via the internet.
The doctor will analyze the image data using the application which is installed in
his mobile, and this belongs to the second phase of the application.
Precision agriculture makes rapid changes in the agricultural sector, and numerous
organizations are leveraging this technique around the world. IoT makes agriculture
field highly efficient. CropMetrics is a precision agricultural organization con-
centrated on ultra-modern agronomic solutions. Wireless sensor network (WSN)
and wireless moisture sensor network (WMSN) are used in IoT. WSN technology
contains many sensors, which are used to monitor and control to do proper irrigation.
This optimization will maximize the profitability of irrigated crop fields, improves
yields, and increases water usage efficiently. Agricultural drones are applied to
farming in order to help increase crop production and monitor crop growth. Through
the use of advanced sensors and digital imaging capabilities, farmers are able to
use these drones to help them gather a richer picture of their fields. An advantage
of drones is crop health, convenient to use, and increase in yield production. IoT
reduces the human intervention in farming. Today’s sophisticated commercial farms
have exploited advanced technologies; however, IoT introduces more access to
deeper automation and analysis.
In [14], the authors have projected the problems faced by the farmers in the
real world. The farmers will be holding the agricultural land which is measured
in hectares, and it is highly difficult for the farmers to identify the infected crop that
is spread out in a vast area. Besides these criteria, the farmers also face certain other
issues like rainfall prediction, climate, availability of water, and so on. Also, the
farmers are not ready to cultivate different or hybrid crops in their land, as they lack
knowledge and less experience in this scenario. In order to support the farmers in
these conditions, a sensor device can be placed in the soil that is used for monitoring
the climatic changes, temperature, and even soil moisture. The data observed by the
5 Applications of the Internet of Things with the Cloud Computing. . . 87
sensors are stored with the help of cloud storage; this stored information will be used
in future for analyzing and predicting the soil status and also about the feasible crop
which can be cultivated with the existing soil and weather conditions. This concept
of implementing cloud will aid in the expansion of data storage, and this will greatly
reduce the production cost.
IOT is an innovative technology that provides modern and automated poultry farm-
ing. Basic environment parameters for poultry farming are temperature, humidity,
ammonia gas, and water level. For instance, the temperature is controlled by cooling
fan, humidity is monitored by the exhaust fan, ammonia gas is examined by
ventilation window open, and the water level is maintained by DC motor. These
actions are performed by the sensor, and monitoring is done the whole day. These
are done mainly for the growth and health of the chicken. Farming challenges caused
by population growth and climate change have made it one of the first industries to
utilize the IoT in this sector.
The major issue in the secured environment is to safeguard our buildings, assets, and
so on [8]. The basic alarm facilities of earlier days had high visibility and later on
introduced a monitoring camera which has some less scope of treatment. Internet of
Things helps to create smart homes and smart cities by adding a feature of real-time
monitoring services to the environment which concerns more about the security by
the use of smart surveillance and security [6].
To make it simpler, let us consider that the user uses network cameras to
monitor the area under surveillance and high-resolution cameras for the places of
large crowd movement. IoT will offer a greater solution to such disparate things
by considering the above scenarios by combining the video surveillance cameras
and smoke detectors into a “small and single glass pane.” As all the devices will
be monitored over a network, they could share data among them and could take
decisions by themselves by which the security will be enhanced.
The images produced by the video surveillance cameras should have high clarity,
and this would be true by adopting Internet of Things. Nowadays, megapixel
technology has grown to its peak by incorporating them with low-lighting cameras
which are now blooming successor of the present market. These are all concerned
with the wide dynamic range (WDR) of cameras. IoT offers a resolution of at least
three times greater, and even 4 k videos could be generated using IoT.
88 U. Ram Jagannath et al.
5.3 Conclusion
There are many benefits that arise with the increased use of Internet of Things. IoT is
used in our everyday life as it greatly reduces the cost and power. Hence, this chapter
has given an overview of applications using IoT. IoT has the potential to drive
integrated solutions that can make a difference. The entire world is transforming
slowly with the use of IoT in daily life. Also, the people are getting ample
opportunities to enhance their commitments and perform various activities using
smart things. This chapter has given a short description of various applications that
are implemented with IoT and also certain challenges faced in those applications.
Research work can be extended in the challenges faced and giving increased security
to the applications to provide the better, smart and secured environment.
References
1. L. Da Xu, W. He, S. Li, Internet of things in industries: a survey. IEEE Trans. Ind. Inf. 10(4)
(2014)
2. K. Ashton, That ‘internet of things’ thing. RFID J. 22 (2009)
3. D.Y. Joo, J.K. Kim, Creative & active convergence model of IoT. Korea Inst. Ind. Econ. Trade
Korea 1 (2014)
4. A. Gauer, Smart city architecture and its applications based on IoT. Procedia Comput. Sci. 52,
1089–1094 (2015)
5. A. Bagula, L. Castelli, M. Zennaro, On the design of smart parking networks in the smart cities:
An optimal sensor placement model. Sensors 15(7), 15443–15467 (2015)
6. K. Sohraby, D. Minoli, T. Znati, Wireless Sensor Networks: Technology, Protocols, and
Applications (Wiley, Hoboken, 2007)
7. H. Sundmaeker, P. Guillemin, P. Friess, S. Woelfel, Vision and Challenges for Realizing the
Internet of Things (Publications Office of the European Union, Luxembourg, 2010)
8. B. Khoo, RFID as an enabler of the internet of things: issues of security and privacy, in Internet
of Things (iThings/CPSCom), 2011
9. Y. Zhang, Technology framework of the internet of things and its application, in Electrical and
Control Engineering (ICECE), 2011
10. Y. Cao, W. Li, J. Zhang, Real-time traffic information collecting and monitoring system based
on the internet of things, in Pervasive Computing and Applications (ICPCA), 2011
11. A.R. Jaladi, K. Khithani, P. Pawar, K. Malvi, G. Sahoo, Environmental monitoring using
wireless sensor networks (WSN) based on IOT. Int. Res. J. Eng. Technol 4 (2017)
12. J. Bugeja, A. Jacobsson, P. Davidsson, On privacy and security challenges in smart connected
homes, in European Intelligence and Security Informatics Conference, 2016
13. Z.H. Chan, E.L.M. Su, Y.M. Lim, E.E. Langeh Randin, C.F. Yeong, M.A. Abdul Razak, I.
Ariffin, Cellco: portable device for automated blood cell count and abnormal cell detection. J.
Telecommun. Electron Comput Eng 9(3–9) (2017)
14. B.B. Prahlada Rao, P. Saluja, N. Sharma, A. Mittal, S.V. Sharma, Cloud computing for internet
of things & sensing based applications, in 6th International Conference on Sensing Technology
(ICST – 2012)
15. P. Satpute, A. Mohanpukar, Cloud centric IoT framework for supply chain management. Int. J.
Comput. Appl. 118(15) (2015)
16. S.R. Basavaraju, Automatic smart parking system using internet of things (IoT). Int. J. Sci.
Res. Publ. 5(12) (2015)
5 Applications of the Internet of Things with the Cloud Computing. . . 89
17. J. Rajaleksmi, G.G. SivaSankari, IoT framework for smart home using cloud computing via
open source mobile platform. Int. J. Comput. Eng. Appl. (2016), ICCSTAR-2016, Special
Issue, May 2016
18. M. Kasmi, F. Bahloul, H. Tkitek, Smart home based on internet of things and cloud computing,
in 7th International Conference on Sciences of Electronics, Technologies of Information and
Telecommunications (SETIT’16), 2016
Chapter 6
Software-Defined Internet of Things
to Analyze Big Data in Smart Cities
6.1 Introduction
With the use of the ICT, which is one of the fascinating technologies, not only people
are connected with each other but also now things are connected with everything.
Accessibility to all kind of services is made so easy. Health, transportation, emer-
gency response, shopping, utilities, economy, weather, etc., are the main services
which are available through the ICT. Some researchers, including the authors of
this chapter, referred these services as smart services. Smart cities concept also
originated because of advancements in ICT research and inventions. The generation
of data is very high these days as IBM reports that approximately 90% of the data
is generated in the last 2 years [1]. Many research organizations have now started
research on big data and related topics because of this fact. Big data consists of
large sets of data which can be mainly identified through “5Vs” (initially, it was
referred to as “3Vs”; two have been added recently): volume (size of dataset), variety
(range of data type and source), velocity (speed of data in and out), value (how
useful the data is), and veracity (quality of data) [2]. Because of its complexity in
processing, big data becomes more challenging. The conventional network obtained
great success by adopting a hierarchical paradigm. Anyway, the closed systems
of network devices in smart cities, we have to configure many devices with high
complexity when business requirements change. The researchers also cannot deploy
new protocols in the real environment especially when we are focusing on a plethora
of embedded devices.
It has been expected that the internet growth will increase to very high end, that
is, 1.6 × 1021 B [3]. In order to cope with this challenge, a greater bandwidth will be
required by users so they can access any services with good speed. The stability of
network will also be required. Almost half decade ago, the abovementioned issues
led Nick McKeown to present Open Flow [4] and based on Open Flow, the concept
of software-defined networking (SDN) was presented in [5]. Where SDN is based on
the idea of detaching the control plane from the forwarding plane, to break vertical
integration, and to introduce the dynamic network and with the ability to program
it. The SDN allows logical centralization of feedback control, and decisions are
made by the “network brain” with a global network view, which eases network
optimization. The data plane elements become highly efficient and programmable
in the SDN system with packet forwarding devices, and the control plane elements
are represented by a single entity, the controller. In conventional networks, it is
much easier to develop and deploy applications in SDN. Also, with the global view
in SDN, it is straightforward to enforce the consistency of network policies. SDN
represents a major paradigm shift in the evolution of networks, introducing a new
pace of innovations in networking infrastructure.
6.2 A Brief Overview of the SDN and Big Data Roles in the
Development of Smart Cities
Three main advantages of the SDN architecture are as follows: (1) The SDN’s
open architecture guarantees the centralized control and automatic management of
networks. SDN allows network engineers to design, deploy, operate, and maintain
networks using a centralized SDN controller rather than configure a lot of hetero-
geneous devices. (2) The SDN can provide various open APIs to flexible program
networks. The network operating system and network applications can be deployed
on servers that adopt X86 architecture and can control data forwarding by Open
Flow. (3) The SDN architecture decouples the data plane and the control plane by
using Open Flow and virtualized networks. A network becomes a logical resource
that can be configured through software. The core idea of SDN has been used in the
field of routers to build an open, flexible, and modularized reconfigurable router [6].
While some excellent work has been done on big data and SDN, these two important
areas have conventionally been addressed separately in most of the previous works.
However, on the one hand, SDN, is an important networking paradigm, will have
a significant impact on big data applications. Several good features, for example,
separation of the control and data planes, logically centralized control, global view
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 93
of the network, and ability to program the network, are associated with it. For
example, big data is usually processed in cloud data centers. Compared to traditional
data centers, SDN-based data centers can have better performance by dynamically
allocating resources in data centers to different big data applications to meet the
service-level agreements (SLAs) of these big data applications [7–19].
On the other hand, big data is an important network application which has a
profound impact on the design and operation of SDN. Specifically, with the global
view of the network, the logically centralized controller in SDN can obtain big data
from all the different layers (i.e., from physical to application layers) with arbitrary
granularity.
Experience shows in cross-layer design that sharing information among different
layers can improve network performance, the network becomes so complex that
conventional methods are unsuitable to design and optimize such networks. Luckily,
big data analytics, which leverages analytical methods to obtain insights from data
to guide decisions, can help the design and operation of SDN, for example, with big
traffic data analytics, it is easier for the controller to perform traffic engineering to
improve the performance of SDN.
This research proposes the integration of programmable devices with IoT (SDN-
IoT) to provide a variety of big data applications for smart cities followed by
the potentials of SDN in building smart cities. The proposed paradigm starts
with data gathering process from various smart city–enabling technologies such as
smart homes, smart grids, intelligent transportation system (ITS), weather forecast
intelligent systems, and so on. After the data aggregation as a second level,
we present a detailed data processing and management (DPM) level where big
data analysis has been used to filter the useful data. We also have evaluated the
performance of our architecture using Hadoop ecosystem. On top of DPM, we have
also identified the applications level along with varying data flow that is supported
by our evaluated big data architecture. In this, we have considered the future
internet architecture recently named as named data networking (NDN) also known
as content-centric networking (CCN) at the application level. This chapter, finally,
provides open issues that provide a roadmap to follow by the active researchers in
the said domain.
This section presents details of the proposed novel layered architecture which
support high-performance computing for SDIoT and also help to design the system
completely. The proposed system architecture consists of four multilevel with the
objective to enhance the current capacity of storing and the speed of processing. At
each layer, various tasks are supported by the reading and writing capabilities.
94 S. Din et al.
A X
A
Aggregator B Aggregator Y
A
C Z
A
Mappers Reducers
Useful
Filtration
information
HDFS HBase
A X
A
Aggregator B Aggregator Y
A
C Z
A
Intelligent
Response
Transportation
System
System
Fig. 6.1 Proposed internet of software-defined system architecture for smart cities
The method of execution flow in the proposed system is done in various levels,
mainly three major levels and two intermediate support levels. Various embedded
units or devices transmit data to the data processing level using standard networking
technologies. During this communication, data is also received by the respective
users who need it for their various jobs. Figure 6.1 shows the flow diagram of the
proposed system architecture.
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 95
In a smart city, few services are considered for data loading and collection pur-
pose, for example, related medical data, public transportation, population, general
facilities, etc. Normally sensors connect these services for the efficient data loading
and collection. The data is communicated between layers through system ways,
that is, data is passed to the upper levels using the first Intermediate level. The
sensed data is transmitted through the aggregator points (AP) which are attached to
sensors. To reduce the congestion, the APs are further classified into three levels,
that is, zone, local, and global APs. These zone levels take the responsibility of
loading high detail information from each unit in the smart city. The local level
AP (LLAP) is the n charge for aggregating sensor data from similar units such
as medical hospital of online health services, transportation data from roads, etc.
Finally, the global level AP (GLAP) eventually collects data from the LLAP and
send it to the SDN core network. The SDN core network uses the SDN controller to
map the traffic received from the GLAP over the network. The controller is designed
to perform many useful functions, for example, to differentiate sensed data based on
the sensor IDs, topology control, optimizes the duty cycle of sensors attached with
each AP, routing decision based on the user requirements. The SDN-enabled router
using a priority application table to determine application-specific routing of data.
In a typical scenario, under the IoT, many thousands of devices are interconnected
for various services to produce high volume data. It is certain that because of the
congestion, the network operation will slow down. The literature shows a number
of methods developed by different researchers to identify congestion level on any
given link. For example, Kandula in [20] developed a method based on a predefined
threshold. It is suggested and proposed that up to 70% capacity used link can be
declared as a congested link. A similar predefined threshold with 75% of link usage
approach is adopted in this research. In order to efficiently transfer big data over
SDN, the design should mention a constraint of this work. Therefore, the current
traffic engineering technique available in the literature has been adopted to enhance
the performance of the SDN. An outstanding suggestion will be given to use/choose
a traffic engineering mechanism for controlling and routing traffic over an SDN. One
example would be to use high-speed processing switches and routers because of the
high volume data generated through the sensors in an IoT environment. The poll-
and push-based methods are used by SDN controller to obtain useful information
about the link (as mentioned above) statistics from switches. It is generally believed
that poll based is a slower method as compared to the push-based mechanisms.
Because of this feature, push-based method can be used for high-speed data routing
and switching. Chen in [21] proposed double-level congestion handling schemes
which are useful to be implemented. Chen in his work proposed that the SDN
controller maintains a global view of the whole network by getting information of
links from each switch in the SDN. This information is further used to control the
load on each link in the network.
96 S. Din et al.
Table 6.1 shows the data aggregation in general, where data of the particular
block is aggregated at different time intervals (e.g., t0 , t1 , . . . tm ) based on
sensorIDs.
FV is a specific attribute called factor variable. The FV is used to represent a
particular attribute of the data block by which the noise is detected and discarded
during data filtration process. The data is in raw and not complied; therefore, in the
first instance the data is sent to the preprocessing stage to the data management and
processing level which are explained as follows:
To convert data to a meaningful form, the data received from the intermediate
level 1 is sent to data processing level for normalization purposes. The normalized
data is useful for the user to make a decision such as a road congestion data is
useful for the commutator to make a decision to reach the destination as quick as
possible and using a short route. As mentioned earlier that high processing power
would require processing high volume data. Conventional computational methods
are always challenged while processing real-time data. The latest technology such
as Hadoop ecosystem and GraphX and SPARK are deployed by the data processing
level in this kind of situation (real-time data processing). Further other capabilities
of the Hadoop system such as HDFS and heterogeneous cluster Hadoop system
are also used for processing, storing, and manipulation tasks related to huge data.
The current literature abounds with many techniques to assign job to the Hadoop
cluster system. In real-time processing, a division of huge jobs to sub job will be
required through a proper scheduling mechanism a map-reduce part. After loading
sub job to map-reduce system, the system allows to alter the number of jobs easily
(if required). Therefore, an adaptive (intelligent) job scheduling technique is used in
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 97
this research to adjust the load on the map-reduce system dynamically. A job tracker
is assigned to each job which uses two different parameters to switch the part of the
job from the current state to another state such as from processing unit to memory
or any other part of the system. The change of sub job from one state to another state
is executed in real-time environment and total based on the amount of a load of a
cluster. This is not a big challenge in Hadoop system environment. However, fixed
job assignment does not produce optimize results heterogeneous Hadoop clusters
systems.
The proposed system (scheduling mechanism) provides a solution to two main
problems because of the fixed job assignment, that is, (1) a high-performance node
remains in the idle state and (2) a low-performance mode always remains in the
busy state. This trade-off between high- and low-performance mode makes the
system unstable for heterogeneous Hadoop clusters. In the proposed system, our
developed job scheduling strategy checks the load on each node at runtime. A node
always automatically demands new jobs, if its current workload is dropped from the
threshold value of 75% of its total link capacity. During each cycle, the proposed
scheduler checks the existing load of each node and loads new jobs accordingly.
Hence, it optimizes the total capacity of a node by incorporating the load parameter
during a single turn. The output of the map-reduce system is also passed to HDFS
system for storing and other required operations.
Rules Definition Various threshold values which are a threshold limit lalue (TLV)
are defined and set for evaluation of the datasets after the completion of the
preprocessing stage. It is understood that TLVs are set for every dataset, for example,
fire detection function, water consumption, etc. Normally TLVs are specific numeric
values to represent some function as discussed above, for instance, 90% pollution,
etc. Along with the TLV, certain rules are also defined to make a decision when that
value is met or surpassed.
Algorithm 6.2: Rules Definition using IF/THEN
Rules Definition
Begin:
IF Traffic_Data > TLV
Traffic Congestion ()
IF Temp > TLV
Fire_Detected
// and so forth
END
Algorithm 6.2 is used to define rules where TLV is the specifically defined
threshold limit value and Temp are the temperatures.
Message Queue After the definition of rules, the rules and data are stored system-
ically in the message queue (MQ). Using MQ mechanism offers some advantages
that do not necessitate an instant reaction to carry on the processing. It also avoids
delay and halts during the processing. The working mechanism of the MQ is very
straightforward by receiving M message at time t; the MQ takes action accordingly.
Figure 6.2 shows the handler H which is handling the overall process in this regard.
98 S. Din et al.
D1
Dn-1
Dn
Dn+1
The proposed architecture has used basic M/M/1 queuing model because of its
efficient functioning wherein M/M/1 model system (e.g., S system) is in a stable
state if coming rate of the messages (data values) is less than the required service
rate. This mechanism, which is used in the proposed architecture, is shown in
Fig. 6.3 with the communication from one data item of the dataset to another. The
mathematical process of the M/M/1 mechanism is illustrated as follows:
D0 and D1 are maintained balance with each other which is required during
steady state, hence for D2 , D3 , . . . Dn−1 , Dn , Dn+1 , such as:
λD0 = μD1
λD1 = μD2
λD2 = μD3
...
λDn−1 = μDn
λ µ
1 2 3
λ λ λ
D0 1 + + + +... =1
μ μ μ
The abovementioned process and equation yield the sum of geometric series,
therefore,
λ
D0 = 1 −
μ
λ
D=
μ
Dn = S n ( 1 − D)
d 1 1
M= ( 1 − D) D =
dD 1−D 1−D
N D 1 1
T = = . =
λ 1−D λ μ−λ
100 S. Din et al.
D D2 λ2
Mq = M − Tasks = .S = =
1−D 1−D (μ − λ)μ
1 1 1 λ
Wq = T − = − =
μ μ−λ μ (μ − λ)μ
D D2
M= =S+
1−D 1−D
The SDN receives data from the HDFS system at the application level via interme-
diate level 2 (IL2). There is no difference between the working of the IL2 and IL1
except for the level of the traffic is different on the SDN is less compared to IL1. The
application level consists of two parts: (1) event and decision management and (2)
named data network (NDN). In the event and decision module, an event is generated
based on the data from the data and processing level. The result, that is, the event is
communicated to the concerned departments to be processed for the respective user.
The management module divides generated event into two groups, that is, high-
and low-level events. As it is clear from the name that the high-level events are
the most important events and processed on a priority basis where the other (low-
level events) are saved in the decision module until a notification is sent back to
the data processing level. After the proper acknowledgment and notification by the
various required modules such as decision management module, it is processed and
discarded. To fully understand the process of the event and decision module, it is
explained through a comprehensive scenario.
Consider an example of a situation of the city where the city collects the data
of a road congestion level. The data is communicated to the event and decision
module. If the communicated data is more than the level of the predefined threshold
which will be considered as a high-level event by the decision module and relevant
messages are sent to the concerned department. On the other hand (e.g., the
congestion level is less than the threshold value), the event is discarded after proper
notification. During this process, all users will also be informed about the situation
by the decision module as discussed earlier. Such transmission of data is performed
to either fulfill user request or automatically broadcast to a group of users following
the hierarchal model presented in Sect. 6.1. In both situations, ICN-based networks
are preferred to send it to the user considering user’s interest [22]. Therefore, a good
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 101
way to use a named data network (NDN) to efficiently fulfill the user requirements
either by sub/pub or pull-based communications as which is used in this research.
Every decision module performs as an NDN node which consists of three entities,
that is, pending interest table (PIT), content store (CS), and a forward information
base (FIB). To avoid interest looping problem, PIT keeps the pending interests and
their unique nonce values. In research presented in [23], it is performed using CS
and FIB by storing incoming content and routing of content to another decision
module. An interest packet sends it to the NDN whenever a user is interested in
a particular data. The interested data is processed, and content is delivered from a
decision module following the NDN.
As mentioned earlier that Spark and GraphX with Hadoop single node setup on
UBUNTU 14.04 LTS coreTMi5 machine with 3.2 GHz processor and 4 GB memory
has been used to implement the proposed system. Pcap packets are generated from
the datasets by using Wireshark libraries and retransmit them towards the developed
system. Hadoop-pcap-lib, Hadoop-pcap-side, and Hadoop Pcap Input libraries have
been considered for network packets processing and generating Hadoop Readable
format collection and aggregation unit so it can be processed easily by these systems
(Hadoop and GraphX). To make smart transportation decisions, GraphX is used.
The data has been taken from [9, 10, 24]. Varieties in the traffic intensities at various
timings on the same road and the intensity analysis at various times of the day
provide great support to the authorities to manage, plan, and monitor traffic on any
particular time.
The intensity of the traffic of one of the roads of Aarhus city is shown in Fig. 6.4.
It indicates that the morning hours between 7:00 and 9:00 and noon timing from
11:25 to 12:30, the traffic is heavy on the roads. The heavy traffic may be due to
the school start and end timing and office start timing. It is a good performance
by the proposed system which announces such a situation and issue alerts. It not
only announces the heavy traffic but also looks into the blockage at roads if any.
Various methods have been used to define the blockage which includes the number
of vehicles and the average speed. If the average speed is less than a specific value
(threshold), then it can be predicted that there is some sort of blockage at the road.
The same information is communicated to the concerned users where they will try
to find the time at which there is very less traffic.
As an example, the analysis is performed on Aarhus city traffic, and the analysis
of the speed on the intensity of traffic is shown in Fig. 6.5. When the intensity of
traffic is more, that is, more vehicles on the road between two points, the average
speed of the vehicles is greater. The fall in some vehicles on the road results in a
rise in the average speed. It can be easily noticed that in a high number of cars (in
between 25 and 30), the average speed is very low at various times of the day, shown
102 S. Din et al.
14
12
10
No. of vehicals
8
6
4
2
0
07:50:00
08:25:00
08:55:00
09:25:00
09:55:00
10:25:00
10:55:00
11:25:00
11:55:00
12:25:00
12:55:00
13:25:00
13:55:00
14:30:00
15:05:00
15:35:00
16:15:00
17:35:00
20:55:00
16:55:00
23:25:00
05:35:00
06:15:00
06:50:00
Time
80
Vehicular Count : 0-10 Vehicular Count : 11-20
70 Vehicular Count : 21-25 Vehicular Count : 25-30
60
Average Speed (km/h)
50
40
30
20
10
0
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 t12 t13 t14 t15 t16 t17 t18 t19 t20 t21
Time
as a red color graph. In case of the low intensity in a number of cars (from 0 to 10)
shown as a blue color, the average speed of the vehicles is quite higher.
The data shows some abnormalities with the less number of cars and other
vehicles, and these abnormalities happen because of other reasons such as road
is under construction or traffic is diverted to some road because of the incident.
Usually, the distance is conserved to measure the time to reach the destination with
an observation that the number of cars and speed (average) also affects the total time
required to reach the target destination.
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 103
16 20
14
12 15
10
8 10
6
4 5
2
0 0
2014-10-01T05:40:00
2014-10-01T05:55:00
2014-10-01T06:10:00
2014-10-06T05:55:00
2014-10-08T05:35:00
2014-10-09T05:55:00
2014-10-22T06:00:00
2014-10-23T05:45:00
2014-10-23T06:00:00
2014-10-24T05:30:00
2014-10-27T07:10:00
2014-10-28T06:35:00
2014-10-28T06:55:00
2014-10-28T07:15:00
2014-10-29T06:35:00
2014-10-30T06:55:00
2014-11-03T06:30:00
2014-11-03T06:50:00
2014-11-04T06:45:00
2014-11-04T07:05:00
2014-11-06T06:45:00
2014-11-06T07:00:00
2014-11-10T06:35:00
2014-11-10T06:55:00
2014-11-11T07:05:00
2014-11-12T06:50:00
2014-11-13T06:20:00
2014-11-13T07:05:00
2014-10-02T06:00:00
2014-10-27T06:55:00
2014-11-05T06:50:00
2014-11-13T06:45:00
Date and Time
Figure 6.6 shows the road blockage in the city. Based on the proposed scheme,
the average speed of the vehicles is too low even when there are a minimum number
of vehicles on the road. It can be seen from various graphs that the road blockage
is shown in the morning hours on different days. One of the main reasons for this
is the construction work on the roads in the morning hours. In such cases, one can
take real-time traffic information to calculate the shortest and quickest path between
source and destination rather than only the distance information.
The humidity at home, shown in percentage in Fig. 6.7, plays a vital role in
user behavior while doing some physical exercise or work. Furthermore, increase
in the humidity demands more resources such as the usage of electricity, etc. The
proposed system exploits this phenomenon through a learning mechanism where the
sensor devices transferred data to the proposed scheme for experiencing the level of
humidity. The proposed scheme in December (2016) takes into considerations some
readings from the sensors and made threshold value. Using the past knowledge, the
proposed system starts predicting the humidity level for January (2016) and others.
The prediction of humidity levels provides great support to users to decide once
they know the level of the humidity as shown in Fig. 6.7. Figure 6.8 shows a
similar data but outside temperature. The processing of large graphs to achieve smart
transportation is the main contribution of this research work; the system is evaluated
on efficiency regarding throughput (in megabytes/sec Mbps) and the response time
(in milliseconds).
104 S. Din et al.
100
Humidity (%)
80
60
40
20
0
0 5 10 15 20 25 30
Date (January 2016)
0 5 10 15 20 25 30
20
15
Temperature (oC)
10
-5
-10
-15
Date (January 2016)
High Avg Low
6.5 Conclusion
In this research, a novel technique is presented to analyze and process big data
for useful application in a smart city environment. The latest technologies such as
Hadoop ecosystem alongside Spark and GraphX are deployed while developing the
abovementioned system. The novel technique is based on a three-level architecture
for efficient data gathering in smart city environment using sensor-based devices. To
normalize data for good use, a preprocessing stage is also used where the collected
6 Software-Defined Internet of Things to Analyze Big Data in Smart Cities 105
data is passed to the upper levels for processing using SDN and NDN. APs are used
to get huge data and pass it to Hadoop ecosystem. A novel scheduling mechanism
is implanted in the Hadoop ecosystem to balance the load on the Hadoop ecosystem
efficiently. A comprehensive decision module is used, which performs and makes
different decisions using the predefined threshold values. The user can also request
in real time in the smart city environment to get support to make a decision.
A simulation test has been implemented for the proposed system on various
data obtained from authentic sources. The simulation results indicate that the
proposed system produced encouraging results. The use of Hadoop ecosystem was
also outstanding and significant by analyzing the data with better speed and high
throughput. Implanting cluster-based Hadoop system with sophisticated scheduling
mechanisms to further reduce the processing will be one of the major tasks for future
of this proposed research.
Acknowledgment This study was supported by the National Research Foundation of Korea
(NRF) grant funded by the Korean government (NRF-2017R1C1B5017464).
References
1. IBM, Armonk, NY, USA. Four vendor views on big data and big data analytics: IBM [Online],
2012. Available: http://www-01.ibm.com/software/in/data/bigdata/
2. V.N. Gudivada, R. Baeza-Yates, V.V. Raghavan, Big data: promises and problems. Computer
48(3), 20–23 (2015)
3. C. V. N. Index, Forecast and methodology, 2013–2018, 2013
4. N. McKeown et al., OpenFlow: enabling innovation in campus networks. ACM SIGCOMM
Comput. Commun. Rev. 38(2), 69–74 (2008)
5. N. McKeown, Software-defined networking. INFOCOM Keynote Speech 17(2), 30–32 (2009)
6. K. Xu et al., Toward a practical reconfigurable router: a software component development
approach. IEEE Netw. 28(5), 74–80 (2014)
7. F. Al-Turjman, Fog-based caching in software-defined information-centric networks. Comput
Electr Eng J 69(1), 54–67 (2018)
8. Y. Han et al., Software defined networking-based traffic engineering for data center networks,
in Proc. 16th Asia-Pacific Network Operations and Management Symposium, Taiwan, Septem-
ber 2014
9. S. Kolozali, M. Bermudez-Edo, D. Puschmann, F. Ganz, P. Barnaghi, A knowledge-based
approach for real-time IoT data stream annotation and processing, in Proc. of the 2014 IEEE
International Conference on Internet of Things (iThings 2014), Taipei, Taiwan, September
2014
10. M. Gramaglia, O. Trullols-Cruces, D. Naboulsi, M. Fiore, M. Calderon, Vehicular networks on
two Madrid highways, in 2014 Eleventh Annual IEEE International Conference on Sensing,
Communication, and Networking (SECON), Singapore, 3 July 2014, pp. 423–431
11. A. Ahmad, A. Paul, M. Rathore, H. Chang, An efficient multidimensional big data fusion
approach in machine-to-machine communication. ACM Trans. Embed. Comput. Syst. (TECS)
15(2), 39 (2016)
12. J. Rowley, The wisdom hierarchy: representations of the DIKW hierarchy. J. Inf. Sci. 33(2),
163–180 (2007)
106 S. Din et al.
7.1 Introduction
Cloud computing concept has been built with the assumptions of good Internet
connectivity, adequate bandwidth and low latency. But with the proliferation of
Internet of Things (IoT) of resource-constrained smart devices, stringent Internet
connectivity demand, high bandwidth, low latency, lower energy consumption,
context-awareness, mobility and enhanced security requirements placed on appli-
cations and services [2, 39] in Smart Cities and Industry 4.0 concepts, as well
as the huge amount of traffic and data expected to be generated, the traditional
cloud-centric architectural arrangement no longer holds due to these assumptions.
Cloud computing is therefore gradually evolving into new complementary concepts
as edge and fog computing and also dew and mist computing to address the
concerns mentioned above by extending the capabilities of cloud computing to the
extreme edge of the network closer to the data generation source. Mist computing
extends the concept of fog computing to the extreme edge of the network at the
level of microcontroller and embedded nodes and is based on IoT concept; on the
other hand, dew computing is based on the Internet and client–server concepts,
whereby on-premises computer provides functionality that is independent of cloud
services, but also collaborative with cloud services [59]. The overall objective of
dew computing is to enable access to cloud computing services in the absence of
Internet connectivity.
In this chapter, an overview of mist computing model/architecture is provided
together with the scope and evolving nature of the topic, its characteristics and
comparison to cloud and fog computing, application scenarios, possible challenges
and future direction. There exist several surveys on fog, dew and mist computing,
but this chapter, therefore, does not intend to revisit these concepts in their entirety.
However, this chapter aims to discuss mist computing in Smart Cities and Industry
4.0 in the context of Africa. The study is an extension of the work done in
[19], which focused on how Nigeria and other developing ICT economies in
Africa can benefit from cloud computing and its evolving and complementary
implementations, challenges, drivers supporting its growth and future vision in the
region.
Section 7.2 of this chapter provides overview on the overlapping and com-
plementary key background concepts, namely, IoT, big data, cloud computing,
edge computing and its implementation. The architecture of mist computing,
its characteristics and its similarity and dissimilarity with cloud computing, fog
computing and edge computing is covered in Section 7.3. In Section 7.4, the
potential, application and use case scenarios of mist computing with respect to two
emerging concepts, namely, Smart Cities and Industry 4.0 with focus on Africa, are
discussed. In Sect. 7.5, an outline of the possible challenges in the implementation
of mist computing, based on observation and trends, is discussed together with
the drivers supporting the growth and technical recommendations in the context of
Africa. Section 7.5 also discusses the future perspective of Smart Cities and Industry
4.0, while Sect. 7.6 concludes this chapter.
Several wide-range and inclusive surveys [2, 8, 17] and studies have been
undertaken in recent times on cloud computing [13, 45, 56], edge computing [2],
fog computing [10, 25] and its implementation in Smart Cities [38] and Industry 4.0
[33], mobile edge computing [1, 29, 30, 53], cloudlet [37, 48, 49], dew computing
[44, 52, 59, 60] and mist computing [39, 40].
The web search interest over a period of 12 months as measured by Google search
trends on fog, edge, mist and dew computing is shown in Fig. 7.1. The Google
trend search shows a very low interest and possibly low awareness overall on these
evolving cloud computing concepts, most especially in Africa. This motivates the
essence and purpose of the research work presented in this chapter.
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 109
In this section, an overview of key enabling concepts and technologies that are
highly intertwined and overlapping with mist and dew computing is given as it
relates to Smart Cities and Industry 4.0.
Internet of
Internet of
Things Com-
munication
people
Technology
Embedded
Application
Device
IoT-created data will be stored, processed, analysed and acted upon close to/at the
edge of the network.
The annual IoT Forum Africa (IoTFA), a platform that enables industry and
expert practitioners to share their vision on IoT trends, challenges and solutions
in virtually all aspects of life and in different industries, reinforces how IoT has the
potential of transforming the African continent.
The rate of deployment and expansion of devices and sensors connected to the IoT
is a major source of data generated and transmitted in real time. Cisco projects
that by 2020, the number of connected devices will exceed 50 billion [21], while
some experts have even projected higher numbers due to the rapid advancement
in the Internet and Internet device technology. It must be mentioned that with
respect to these generated data, cloud computing facilitates storage, processing
and analysis. Overtime data collection and analytics have evolved into three (3)
ways – Analytic 1.0, collection of historical data; Analytic 2.0, social media and
unstructured information; and Analytic 3.0, real-time IoT data from a vast number
of sensors (heterogeneity) – making it challenging to push data to a single cloud
data warehouse, hence the need to bring data analytics closer to the edge of network
into routers and gateways, as well as on-board embedded systems with sensors (mist
computing).
The characteristics of big data are captured in the 5 V’s model, namely,
volume, variety, velocity, veracity and value [35]; hence, a data that meets these
characteristics is termed big data. Big data in Africa has numerous applications,
such as in climate change, poverty and disease surveillance, agriculture, banking
and finance, supply chain, media, space research and biological research [34, 41,
54].
Edge computing (EC) has evolved from cloud computing due to the tremendous
growth in IoT devices, which according to Cisco is estimated to reach 50 billion
devices in no distant future. This comes with several challenges to the traditional
cloud computing arrangement on several fronts such as the need for low latency,
high bandwidth, enhanced security due to several organizations’ sentiments on their
data residing in unknown third party’s data centre and the growing demand and
advocacy for a greener technology. Hence, the need to extend cloud computing
capabilities closer to the data source on the edge of the enterprise network
using the edge gateway’s core routers and switches is necessary. Overtime, edge
computing has also evolved into other complementary implementations to address
challenges and flaws associated with cloud and edge computing. A brief overview
of the technological implementations of edge computing is discussed over the next
subsections. For a more detailed survey, readers are referred to [8].
Fog computing (FC) was originally proposed by Cisco [10], which extends cloud
computing to the edge of the network at the level of routers and gateways. Its
emergence is due to developing applications that are sensitive to latency, which
cannot be met by the current cloud-centric architectural arrangement. There is
currently no globally accepted definition of FC; however, for the purpose of the
research work presented in this chapter, the definition formally stated by [10]
is adopted, which states that: “Fog computing is a highly virtualized platform
that provides compute, storage and networking services between end devices and
traditional Cloud Computing data centres, typically, but not exclusively located
at the edge of the network” with the fog nodes as its building blocks which are
“distributed fog computing entities enabling the deployment of fog services and
formed by at least one or more physical devices with processing and sensing
capabilities (e.g., computer, mobile phone, smart edge device, car, temperature
sensors, etc.). All physical devices of a fog node are connected by different network
technologies (wired and wireless) and aggregated and abstracted to be viewed as
one single logical entity, that is the fog node, able to seamlessly execute distributed
services, as it were on a single device” [20].
Figure 7.4 depicts the fog to cloud architecture, with one cloud layer and two
fog layers. Fog layer 1 is directly connected to the edge devices, while fog layer
2 is in between fog layer 1 and the cloud, acting as an intermediate processing
layer. According to the OpenFog Consortium, advantages of fog computing include
enhanced security due to the distributed architecture, cognition/decision-making on
fog node, improved agility, lower latency and overall efficiency.
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 113
CLOUD
Internet
FOG NODE
Fog layer 2
FOG NODE
FOG NODE
Edge devices
Edge devices Fog layer 2
7.2.4.3 Cloudlets
Security, privacy and regulatory frameworks [5] are key concerns relating to new
and emerging technologies in cloud computing, Smart Cities and Industry 4.0,
due to the sensitive information generated and transmitted by resource-constrained
smart terminal devices in the IoT ecosystem which are prone to security attacks.
Blockchain technology, a peer-to-peer (P2P) distributed ledger technology for
transparent transaction devoid of a trusted intermediary, has evolved beyond its
original application to support bitcoin cryptocurrency and is seen as a viable option
to address these concerns by integrating it with smart devices in smart cities in order
to provide a secure communication platform [9] and with cyber-physical systems
(CPS) in Industry 4.0 to provide transparent transactions among smart devices [7].
Mist computing has evolved from fog and cloud computing into a collaborative
cloud technology based on the idea that communication should be made at the
level of sensors and actuators without burden on the communication network and
the Internet, by leveraging on the computational networking resources from the
devices at the very edge of the IoT networks. This leads to increased autonomy of
the system and reduced latency and bandwidth usage with a corresponding power
consumption reduction as communication accounts for five times consumption
compared to computation. “Mist Computing represents a paradigm in which edge
network devices, that have predictable accessibility, provide their computational
and communicative resources as services to their vicinity via Device-to-Device
communication protocols. Requesters in Mist can distribute software processes to
Mist service providers for execution” [28]. Scalability, reconfigurability, location
self-awareness, situation awareness and attention and machine-to-machine com-
munication (M2M) are core features of mist computing [39]. Cisco is leading
the vision to move data analytics closer to where the data is generated for better
decision support of mission critical services instead of backhauling it to the cloud
or even the fog. IoT smart devices should not just be about data collection; most
importantly, they should bring intelligence to the extreme edge of the network
to derive value from the data for quicker decision support. Fog computing has
been able to address a lot of challenges associated with IoT, including bandwidth
and latency requirements. In fog computing, the gateway is key to ensuring the
coordination and functioning of IoT applications; however, this arrangement has
some drawbacks, such as the gateway being the single point of failure since the
116 E. M. Dogo et al.
entire network is dependent on it [40]. For this reason, mist computing is gaining
popularity fast and is seen to address some of the concerns on cloud and fog
computing architectures with Thinnect, an IoT edge network service provider, at
the forefront of the development and implementation of mist computing in real-life
scenarios.
Since the emergence of the concept of dew computing originally proposed in [58],
other technical definitions have also emerged from various researchers [22, 52].
In the context of the research work presented in this chapter, the definition in
Wang (2016) [59] is adopted, which states that: “Dew computing is an on-premises
computer software-hardware organization paradigm in the cloud computing envi-
ronment where the on-premises computer provides functionality that is independent
of cloud services and is also collaborative with cloud services”. The goal of
dew computing is to fully realize the potentials of on-premises computers and
cloud services. The vision behind dew computing is based on the concept of the
Internet, whereby it envisioned users situated in any part of an enterprise network
to access the cloud without the continuous access to the Internet using a client–
server model, in a complementary manner to cloud computing [46]. Figure 7.6
depicts a conceptualized model of dew–cloud architecture comprising of dew server,
dew DBMS, dew client program, dew client service application and high-speed
connection to the cloud server. For more details on the concepts of dew computing,
readers are referred to [44, 59, 60].
A summary of the comparison between cloud, fog and mist computing is shown in
Table 7.1, which is based on earlier studies by [25].
In this section, the potentials, applications and use case scenarios of mist computing
will be broadly discussed with respect to two emerging conceptual domains, namely,
Smart Cities and Industry 4.0.
118 E. M. Dogo et al.
Table 7.1 Comparison of mist with cloud, edge and fog computing
Cloud Edge
computing computing Fog computing Mist computing
Support for IoT Yes Yes Yes Yes
Infrastructure Centralized and Centralized Decentralized Centralized/distributed
deployment accessed via the LAN or IoT gate- with the
model internet way/distributed microcontroller
network
Efficiency High Higher Higher Higher
Latency High 100 ms Low Low 10–100 ms Ultra-low 10–100 ms
and above
Bandwidth High Low Low Low 1–300 kbps
1Mbps–10Gbps 100 kbps–1Mbps
Computational High Moderate Moderate Low
power
Mobility support No Yes Yes Yes
of IoT services
Data status Data at rest Data in Data in transit/at Data in transit
transit/at rest rest
Security and Higher concern High Medium concern Low concern since
privacy concerns due to sending data reside with the
raw data over hardware
the internet
QoS Yes Yes Yes Yes (limited)
Geographical Limited spread Limited spread Wide spread in Very wide spread in
availability in hundreds in hundreds to thousand to million–trillion
thousands million
Service type Global Limited Limited Very limited
Hardware Large Limited Limited Very limited
Working Indoor data Indoor Outdoor/indoor Outdoor/indoor
environment centre
Power Very high High Low Very low
consumption
Online/offline Only online Dual Dual online/offline
capabilities online/offline
Addressing IPv4/IPv6 IPv4/IPv6 IPv4/IPv6 IPv6 (6LowPAN)
Cost of Medium Medium High Low
implementation
African cities are vital centres of trade, commerce, innovation and many other
important activities. Recent statistics estimate that by 2030, over 60% of the
African population would be dwelling in cities. In addition to this fact, these cities
approximately contribute 70% of greenhouse gas emissions and energy utilization
while occupying and accounting for only 5% of the continental land mass [14, 26].
120 E. M. Dogo et al.
Apart from these environmental trends, African cities are presently experiencing
an increasingly huge demand for food, water, shelter and building materials, risk
and disaster control strategies, waste management schemes and pollution control
strategies [26].
African cities are therefore under constant and immense pressure to provide
better standard of living, offer quality amenities and services, address social
and environmental issues, foster economic growth and trade competitiveness,
promote regional urbanization and industrial mechanization, attract investments
and reduce costs [26]. These pressures are necessitating African cities to look
into smart solutions in order to tackle the plethora of problems and challenges
on the ground. Such solutions will enable African cities to transform into smart
sustainable cities where ICT innovations and high-tech social amenities are used to
provide improved standard of living, quality services delivery, seamless urbanized
operations, economic self-reliance and growth trends that will match up with the
environmental demands, social requirements and economic needs of the present and
future generations.
Industrial experts and field specialists have shown that mist computing is more
suitable for realizing smart city solutions where streets intelligently adapt to
dynamic events, conditions, situations and changes in the city [15, 16, 39]. An
interesting innovation which can possibly serve as a model for African cities to anal-
yse, adapt and adopt is the smart street light control system developed by Cityntel
[57]. This system is based on next-generation flat wireless mesh network (for direct
application and device layer communications) and mist computing stack where
controllers are equipped with data processing and embedded intelligence features.
These smart controllers carry out network-based and/or device-specific decision-
making operations and rely on wireless communication for data transmission and
reception with sensors and detectors, such as movement and road surface detectors
and noise, light reflection, air pollution, humidity and temperature sensors [57].
Unlike existing conventional systems, this solution replaces centralized perma-
nent remote server control with distributed intelligent adaptive and autonomous
situation- and context-aware devices. In this system, the LED street lights dynam-
ically adjust brightness based on updates and notifications regarding the status of
weather condition, human presence and movement and nature of traffic flow or
intensity. In a situation where there is little traffic, street lights will switch to energy-
saving mode by dimming their lights. Centralized computing is also utilized for
processing aggregated data in order to make results of statistical analysis readily
available for low-level device programming and learning. Energy conservation,
minimized cost, easy upgrade and compatibility, fast and cost-effective deployment
and high precision and reliability are ensured due to the adoption of adaptive,
local communication and distributed computing strategies among low-cost wireless
sensors for desirable system performance [57].
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 121
Despite all the numerous potential benefits associated with the concept of smart
cities, information security and privacy remain a concern to stakeholders due to the
vast amount of vital statistics and transit information generated by interconnected
smart devices in the IoT ecosystem. There is, therefore, a need for information to:
(i) be confidential, non-disclosure of sensitive information; (ii) possess integrity,
information is accurate and reliable; and (iii) not manipulated or corrupted and
available whenever/wherever needed by authorized persons. A lot of research
has been carried out to address cloud security, but challenges remain due to the
heterogeneous nature and resource constraints of smart devices, compatibility and
other pertinent factors. A comprehensive review of cloud security is discussed in
[27]. Recently, Blockchain which is a P2P distributed ledger technology devoid
of a trustworthy intermediary (originally developed to support cryptocurrency) has
received a lot of attention as a solution to address security concerns in numerous
domains including smart cities [9]. Figure 7.7 shows a proposed smart city security
framework across the physical, communication, database and application layers of
IoT device to enable a common framework for secured data communication by
different smart devices in a distributed community of devices.
Industry 4.0 can be conceptually viewed as embracing and infusing the core
ideas of mist computing in the factory with the goal of gaining operational and
performance efficiencies through rapid and precise decision-making for automated
devices in the factory environment [24, 43]. In the context of this research, Industry
4.0 can be technically viewed as the collaborative use of mist computing, smart
environmental sensors, robots, big data and mobile devices via cognitive and
independent automation in order to achieve flexible mass production, easy product
customization, scalability, predictability and greater efficiency during the entire
manufacturing process [42].
This fourth industrial revolution goes beyond simply automating production,
and it is based on cyber-physical systems achieved through complete and seamless
digitalization of the manufacturing industry [24, 42]. Industrial locations and
manufacturing operations are unique in Africa since the continental landscapes
are vast with scattered and sporadic business hubs. In addition to this, many local
corporations and manufacturing industries in Africa have matured to the level of
making effective and productive use of data generating technologies and reactive
data analysis strategies. It must also be added that a large number of manufacturing
organizations in Africa are already aware of the ongoing global trends in Industry
4.0 together with the significant potentials and immense benefits of this explosive
technology.
122 E. M. Dogo et al.
However, in comparison with developed ICT economies of the world, the African
continent is still yet to attain full-blown maturity with respect to adopting, imple-
menting and deriving maximum benefits from harnessing the potentials of Industry
4.0. Moreover, the widespread adoption of smart technologies and ICT innovations
that can rapidly catalyse Industry 4.0 is still at a budding stage in most indigenous
manufacturing industries in Africa. This impediment is a result of the general
reluctance to venture into and invest in new knowledge and novel technologies
by the government and industry of most countries in Africa. This hindrance is
also due to the entrenched business culture among African manufacturers where
cost-minimization measures are highly prioritized and expenditures on innovative
technologies are curtailed to generate maximum profit in the prevailing harsh and
competitive economic environment.
An exception to this is the case of African-based large international organizations
with cross-continental operations where the level of adoption of smart solutions and
innovative technologies is encouraging. From the African perspective, the proper
adoption and judicious usage of Industry 4.0 offers limitless opportunities to tackle
some socio-economic problems and prevalent industrial challenges associated with
supply chain logistics through the development of smart machines, smart factories,
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 123
smart plants, smart products, smart technologies and smart services. This will foster
the development of unique, indigenous, high-tech gadgets, products, devices and
services that can successfully leapfrog and dynamically compete with competitors
in the global market. A suitable and relevant use case model for this research
context is Siemens’ MindSphere, which has the potential of serving as an underlying
platform for mist computing services pertinent to resource optimization, energy data
management and preventive/predictive maintenance [50, 51].
Blockchain technology which was originally developed for Bitcoin and cryp-
tocurrency as aforementioned is also gradually finding interesting and fruitful
applications in Industry 4.0. Supply chain management, smart contract, digital
currencies and tighter cybersecurity are applications of Blockchain technology in
Industry 4.0. The need for confidentiality, integrity and availability of sensor data
is critical to industrial applications and manufacturing systems. Therefore, it is
envisioned that Blockchain technology will catalyse remote machine diagnostics
and machine-to-supplier transactions which will lead to improved spare parts
replacement and overall maintenance practices.
Benefits of PdM
Cost Reduction PdM with IIoT could benefit industry by improving efficiency,
reducing inventory carrying costs, saving material costs, improving the return on
asset (ROA), cutting down the cost of engaging service providers, reducing the cost
of repairs and spare parts purchase and reducing unwanted costs associated with
124 E. M. Dogo et al.
unplanned downtime and frequent maintenance. In addition to this, PdM with IIoT
right on site reduces burden on network and overall cloud cost.
Direct and Impactful Value Creation PdM enables data collection without
additional cost as the IIoT sensors are all located where the source of the problem
is likely to occur. This makes it possible for prompt and rapid remote intervention
in urgent repair and maintenance operations. Resultantly, manufacturing industries
employing PdM achieve higher customer satisfaction and add more value and
superior quality to the products and services offered to their customers due to
their proactive decision-making processes, improved production strategies and pre-
emptive business models.
Secure, Safer and Sustainable Operation By adopting PdM, enhanced data
security is ensured as data is collected and analysed right in the factory instead
of sending to the Internet or external data centre or gateways which are prone
to cyberattacks. This enhancement enables manufacturers to deal with unexpected
events and better reduce health, safety, environmental and quality risks in the entire
manufacturing process. In addition to this, there is the advantage of easily and
safely planning service intervals, achieving real-time analytics and attaining ultra-
low response time during machine operation.
Uptime Improvement and Availability PdM ensures increased uptime and avail-
ability of the machine and equipment due to the reliance on automated intervention
of autonomous and cognitive systems that ensure better machine availability and
service life, prolong lifetime of aging assets, prevent lasting damage to relatively
new equipment and foster quality control and precision manufacturing.
Accurate Predictions and Better Decisions By utilizing PdM, a plethora of new
data sources are constantly made available that reliably and accurately reflect the
real operational status of equipment for better decision-making. In addition to this,
PdM ensures that valuable time is expended on data-driven problem solving and
not wasted on brute-force data collection and validation. Another clear advantage of
PdM in terms of economy and efficiency is that IIoT sensors, data collection, data
analytics and decision support for PdM are all with a local area network (LAN).
Skills Gap and Digital Divide The digital divide between developing ICT
economies in Africa and developed ICT economies is not an overstatement, but
this divide is still present and significantly wide in some regions of Africa. This
digital divide, unfortunately, hinders or delays the rate of technical training, capacity
building and skills acquisition necessary for the successful implementation of mist
computing. In addition to this, the present engineering and ICT curriculum in most
African institutions need reforms that will sufficiently enrich the programmes with
practical skills that will adequately prepare and equip Africans for mist computing
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 125
solutions and other emerging innovative technologies. Presently, there is very little
research work that decisively investigates and quantifies the real impact of this
existing digital divide and skills deficit. Therefore, it is necessary in Africa to
embrace a holistic approach that considers the dual goal of investing in smart
people and smart technologies.
Technology Transfer, Assimilation and Domestication It would be very chal-
lenging to realize the latent benefits of mist computing when solutions are simply
imported from developed ICT economies into Africa as turnkey products and
services. These smart technology solutions need to be meticulously studied, prop-
erly understood and, most importantly, technically adapted to be indigenously
relevant, suitable and effective to tackle the specific development needs of Africa.
This is necessary because economic situation, geographical context and cultural
inclination play an indispensable role in the technology transfer process. Therefore,
governments and corporate organizations in Africa need to adopt mist computing
solutions in accordance with their prevailing regional socio-economic needs. For
instance, adopting a mist computing solution that demands massive data aggregation
from a large wireless sensor network for distributed intelligent transport and traffic
management system will be prohibitively expensive and unsuitable based on the
existing economic realities and social demands in most African countries. In such
a scenario, a localized and simplified version of this smart technology solution
that ingeniously harnesses the pervasiveness of mobile telephone data and other
available cost-effective and easily accessible alternatives will be more suitable to
the context of Africa.
Inclusiveness and Governance Smart technology projects involving mist comput-
ing call for smart governance models in order for such technological enterprise
to thrive in Africa without suffering premature and sudden downfall. The need
to develop and implement this pragmatic governance model in Africa is long
overdue as there is an evident disconnect, disharmony and lack of synergy between
polytechnics, universities, indigenous industries, multinational corporations, civil
societies and local, state and federal governments. Such a smart governance model
will bring together and carry along all parties during the course of executing
the project in order to gain varying perspectives and get robust recommendations
and vital suggestions from experts in the economic, legal, educational, political,
technical, financial, ethical, social and developmental fields. The current governance
model for technology projects in most African countries is still not fully inclusive
due to information isolation which acts as a great obstacle to resource integration
in the process of executing technological projects at both managerial and technical
levels. Therefore, there is a need to reform the existing modus operandi and evolve
smart governance models that will enable data and services from mist computing
solutions to be useful for citizens, accessible for decision-making, readily available
and highly effective in tackling the various developmental needs of Africa.
Failure, Faults and Risk Other challenges hindering the full adoption and imple-
mentation of smart technology projects involving mist computing in Africa are
126 E. M. Dogo et al.
Trainable and Resourceful Young Population One of the biggest drivers encour-
aging the establishment of smart technology projects involving mist computing in
Africa is the large youthful population who are able, hardworking and willing to
learn and be trained on the technical and managerial aspects of these innovative
technologies. These vibrant youths are very much cognizant of the global technolog-
ical trends, and they desire to consume and enjoy the services of the same high-tech
quality products as other consumers in developed ICT economies. Moreover, these
African youths are relatively more entrepreneurial, dynamic, flexible, trendy, open-
minded and computer literate, making it very easy and less costly for them to
be trained on the proper understanding, usage, running, handling, operation and
management of smart technology systems involving mist computing.
7 Taking Cloud Computing to the Extreme Edge: A Review of Mist. . . 127
resources and maximize efficiency gains. This approach will unambiguously and
seriously convince interested local, regional and international investors to provide
necessary financial support for smart technology projects involving mist computing
in Africa.
7.6 Conclusion
Smart Cities and Industry 4.0 are still in their budding stage in Africa. But with
the rapid pace of technological transformation with its envisioned disruption to
businesses and society, it is critical for Africa to join, with the possibility of
catching up technologically with the rest of the world. The current trend suggests
a massive shift away from a centralized cloud architecture towards a distributed
architecture with more computing and intelligent capabilities at the edge or extreme
edge of the network closer to users. Mist computing has the potential to solve
many of the unique socio-economic challenges currently facing Africa in different
spheres of life in a collaborative way with the existing cloud and fog computing
frameworks. This can be pragmatically achieved by leveraging on IoT solutions and
Internet technologies such as smart transportation, smart grid, smart exploration and
mining and smart healthcare. With the adoption of these smart solutions, African
cities and industries will be adequately equipped with technologies to respond to
perennial problems in normal and emergency situations such as industrial hazards,
natural disasters, conflicts and insurgencies, scarcity and resource constraints,
infrastructural failures and inadequacies, conventional and cybercrime occurrences
and other forms of socio-economic challenges embattling the African region in a
sustainable manner.
Acknowledgements This work was funded by the Institute for Intelligent Systems and the
Department of Electrical and Electronic Science at the University of Johannesburg, South Africa.
References
1. N. Abbas, Y. Zhang, A. Taherkordi, T. Skeie, Mobile edge computing: a survey. IEEE Int.
Things J. 5(1), 450–465 (2018). https://doi.org/10.1109/JIOT.2017.2750180
2. Y. Ai, M. Peng, K. Zhang, Edge computing technologies for internet of things: a primer. Digit
Commun. Netw. 4(2), 77–86 (2018). https://doi.org/10.1016/j.dcan.2017.07.001
3. I.F. Akyildiz, S. Nie, S. Lin, M. Chandrasekaran, 5G roadmap: 10 key enabling technologies.
Comput. Netw. 106, 17–48 (2016). https://doi.org/10.1016/j.comnet.2016.06.010
4. M. Al Amine, K. Mathias, T. Dyer, Smart cities: how 5G can help municipalities
become vibrant smart cities. Accenture. (2017), https://www.accenture.com/
t20170222T202102__w__/us-en/_acnmedia/PDF-43/Accenture-5G-Municipalities-Become-
Smart-Cities.pdf. Accessed 15 Mar 2018
130 E. M. Dogo et al.
24. K. Hartung, Computer-on-modules for robotics & industry 4.0 automation. Combined Print
Magazine for the European Embedded Market, pp. 6–8 (2017), http://files.iccmedia.com/
magazines/basoct17/basoct17.pdf. Accessed 14 Mar 2018
25. P. Hu, S. Dhelim, H. Ning, T. Qiu, Survey on fog computing: architecture, key tech-
nologies, applications and open issues. J. Netw. Comput. Appl. 98, 27–42 (2017).
https://doi.org/10.1016/j.jnca.2017.09.002
26. J-M. Huet, Smart cities: the key to Africa’s third revolution. BearingPoint (2016), https://
www.bearingpoint.com/files/smart-cities-the-key-to-africas-third-revolution.pdf. Accessed 22
Feb 2018
27. R.L. Krutz, R.D. Vines, Cloud security: A Comprehensive Guide To Secure Cloud Computing
(Wiley Publishing, Indianapolis, 2010)
28. M. Liyanage, C. Chang, S. N. Srirama, mePaaS: Mobile-embedded platform as a service for
distributing fog computing to edge nodes. 2016 17th International Conference on Parallel and
Distributed Computing, Applications and Technologies (PDCAT), Guangzhou, China, 16–18,
pp. 73–80 (Dec 2016). https://doi.org/10.1109/PDCAT.2016.030
29. P. Mach, Z. Becvar, Mobile edge computing: a survey on architecture and
computation offloading. IEEE Commun. Surv. Tutorials 19(3), 1628–1656 (2017).
https://doi.org/10.1109/COMST.2017.2682318
30. Y. Mao, C. You, J. Zhang, K. Huang, K.B. Letaief, A survey on mobile edge computing:
the communication perspective. IEEE Commun. Surv. Tutorials 19(4), 2322–2358 (2017).
https://doi.org/10.1109/COMST.2017.2745201
31. R. Minerva, A. Biru, D. Rotondi, Towards a definition of the internet of things
(IoT). IEEE Internet Things, Revision 1 (2015), https://iot.ieee.org/images/files/pdf/
IEEE_IoT_Towards_Definition_Internet_of_Things_Revision1_27MAY15.pdf. Accessed 14
Mar 2018
32. J. Mundy, 5G in the smart city (2018), https://5g.co.uk/guides/5g-smart-city/. Accessed 12 Mar
2018
33. Nebbiolo Technologies, Fog computing: keystone of industry 4.0 (2017), https://
www.nebbiolo.tech/wp-content/uploads/Nebbiolo-Technologies-solutions-brief.pdf. Accessed
10 Mar 2018
34. N. Pokhriyal, W. Dong, V. Govindaraju, Big Data for Improved Diagnosis of Poverty: A Case
Study of Senegal (The Brookings Institute, Washington, 2015), https://www.brookings.edu/
blog/africa-in-focus/2015/06/02/big-data-for-improved-diagnosis-of-poverty-a-case-study-of-
senegal/. Accessed 4 Apr 2018
35. P.C. Neves, J. Bernardino, Big data in the cloud: a survey. Open J. Big Data (OJBD) 1(2), 1–18
(2015)
36. Nexcom, Industry 4.0 smart services dawn with predictive maintenance (2015),
http://www.nexcom.com/news/Detail/industry-4-0-smart-services-dawn-with-predictive-
maintenance?preview=1. Accessed 14 Mar 2018
37. Z. Pang, L. Sun, Z. Wang, E. Tian, S.Yang A survey of cloudlet based mobile computing. 2015
International Conference on Cloud Computing and Big Data (CCBD), Shanghai, pp. 268–275
(2015). https://doi.org/10.1109/CCBD.2015.54
38. C. Perera, Y. Qin, J. Estrella, S. Reiff-Marganiec, A. Vasilakos, Fog computing for
sustainable smart cities: a survey. ACM Comput. Surveys (CSUR) 50(3), 43 (2017).
https://doi.org/10.1145/3057266
39. J.S. Preden, K. Tammemae, A. Jantsch, M. Leier, A. Riid, E. Calis, The benefits of self-
awareness and attention in fog and mist computing. IEEE Comput. Soc. Comput. 48(7), 37–45
(2015)
40. J. Preden, Evolution of mist computing from fog and cloud computing THINNECT (2014),
http://www.thinnect.com/static/2016/08/cloud-fog-mist-computing-062216.pdf. Accessed 15
Mar 2018
41. I. Protopop, A. Shanoyah, Big data and smallholder farmers: big data applications in the agri-
food supply chain in developing countries. Int. Food Agribusiness Manag. Rev. 19(A), 173–190
(2016)
132 E. M. Dogo et al.
42. J.C. Ramirez, Fog, mist and hardware in factory automation (2017), https://
www.smartindustry.com/blog/smart-industry-connect/fog-mist-and-hardware-in-factory-
automation/. Accessed 14 Mar 2018
43. J.C. Ramirez, Hardware for industrial IoT fog and mist computing. Combined print magazine
for the European Embedded Market, pp. 24–25 (2017), http://files.iccmedia.com/magazines/
basoct17/basoct17.pdf. Accessed 14 Mar 2018
44. P.P. Ray, An introduction to dew computing: definition, concept and implications. IEEE Access
6, 723–737 (2018). https://doi.org/10.1109/ACCESS.2017.2775042
45. B.P. Rimal, E. Choi, I. Lumb, A taxonomy and survey of cloud computing. 2009
Fifth International joint conference on INC, IMS and IDC, Seoul, pp. 44–51 (2009).
https://doi.org/10.1109/NCM.2009.218
46. A. Rindos, Y. Wang, Dew computing: The complementary piece of cloud comput-
ing. 2016 IEEE International conferences on big data and cloud computing (BDCloud),
Social computing and networking (SocialCom), Sustainable computing and communica-
tions (SustainCom) (BDCloud-SocialCom-SustainCom), Atlanta, GA, pp. 15–20 (2016).
https://doi.org/10.1109/BDCloud-SocialCom-SustainCom.2016.14
47. R. Saracco, Industry 4.0 leveraging on 5G (2017), http://sites.ieee.org/futuredirections/2017/
09/19/industry-4-0-leveraging-on-5g/. Accessed 12 Mar 2018
48. M. Satyanarayanan, P. Bahl, R. Caceres, N. Davies, The case for VM-based
cloudlets in mobile computing. IEEE Pervasive Comput. 8(4), 14–23 (2009).
https://doi.org/10.1109/MPRV.2009.82
49. M. Satyanarayanan, Z. Chen, K. Ha, W. Hu, W. Richter and P. Pillai, Cloudlets: at the leading
edge of mobile-cloud convergence. 6th International conference on mobile computing, applica-
tions and services, Austin, pp. 1–9 (2014). https://doi.org/10.4108/icst.mobicase.2014.257757
50. Siemens Discover MindSphere (2018), https://siemens.mindsphere.io. Accessed 14 Mar 2018
51. Siemens MindSphere, the internet of things (IoT) solution (2018), https://www.siemens.com/
global/en/home/products/software/mindsphere.html. Accessed 8 Mar 2018
52. K. Skala, D. Davidovic, E. Afgan, I. Sovic, Z. Sojat, Scalable distributed computing hierarchy:
cloud, fog and dew computing. Open J. Cloud Comput. 2(1), 16–24 (2015)
53. T. Taleb, K. Samdanis, B. Mada, H. Flinck, S. Dutta, D. Sabella, On multi-access edge
computing: a survey of the emerging 5G network edge cloud architecture and orchestration.
IEEE Commun. Surv. Tutorials 19(3), 1657–1681 (2017)
54. R. Taylor, B. Frank, Big data opening R&D opportunities in Africa (2017), http://
www.scidev.net/index.cfm?originalUrl=/sub-saharan-africa/data/opinion/big-data-opening-
opportunities-africa.html&. Accessed 4 Apr 2018
55. A. Varghese, D. Tandur Wireless requirements and challenges in industry 4.0. 2014 Inter-
national conference on contemporary computing and informatics (IC3I), Mysore, India, pp.
634–638 (2014). https://doi.org/10.1109/IC3I.2014.7019732
56. B. Varghese, R. Buyya, Next generation cloud computing: new trends and
research directions. Futur. Gener. Comput. Syst. 79(3), 849–861 (2018).
https://doi.org/10.1016/j.future.2017.09.020
57. A. Vork, J. Tamm LED Lights with smart controllers. SmartEnCity project, Estonia,
Cityntel (2016), https://smartencity.eu/about/solutions/led-lights-with-smart-controllers-tartu/
. Accessed 12 Mar 2018
58. Y. Wang, Cloud-dew architecture. Int. J. Cloud Comput. 4(3), 199–210 (2015)
59. Y. Wang, Definition and categorization of dew computing. Open J. Cloud Comput. 3(1), 1–7
(2016)
60. Y. Wang, Y. Pan, Cloud-dew architecture: realizing the potential of distributed database systems
in unreliable networks. 21th International conference on parallel and distributed processing
techniques and Applications (PDPTA’15), pp. 85–89 (2015)
Chapter 8
IoT and Edge Computing as a Tool
for Bowel Activity Monitoring
Umit Deniz Ulusar, Erdinc Turk, Ahmet Sefa Oztas, Alp Erkan Savli,
Guner Ogunc, and Murat Canpolat
8.1 Introduction
The Internet of Things (IoT) is enabling several applications ranging from envi-
ronment to healthcare and medical fields [1, 2]. The main idea is the ubiquitous
presence around us with a variety of things or objects that are connected to the
Internet. A key application for the IoT-based technologies is wireless medical sensor
networks (WMSNs), which aim to observe patients’ health status remotely without
intervening his/her daily life and perform early diagnosis of possible diseases. For
this, researchers have been developing tiny sensors that can observe heartbeat, blood
pressure, temperature, and so on, apply signal processing algorithms, and then share
it through wireless connectivity with medical professionals. WMSNs are also an
indispensable tool for patients with certain disabilities and accessibility issues by
making remote monitoring possible and minimizing the necessity of hospital visits.
Also, they are beneficial for healthy and active people for monitoring their daily
activities and well-being. A typical example of this is a fall detection device which
can report the situation to an emergency responder.
Unfortunately, so far, the real-world application of IoT-based health monitor-
ing for mobile users suffered from energy constraints imposed by the batteries.
8.2 Background
Auscultation is the typical mean of observing body sounds and is commonly used
in clinical environments. This technology exists since the early 1800s, and despite
all the improvements, the overall aim is to identify specific sounds within the body.
With the advent of technology, IoT-driven bioacoustic sensors can be utilized for
continuous monitoring. There are some studies in the literature that aim to use
bioacoustic sensors for observing pulmonary, heart, and intestinal sounds. Tang et al.
have developed an electronic stethoscope for observing heart sounds with embedded
processor and used Bluetooth for data transmission [3]. Similarly, Rach et al. have
developed a bioacoustic sensor prototype that can operate over long periods and
performed sound analysis [4]. In another research, Patil et al., have developed an
electronic stethoscope to observe heart rhythm [5].
Electronic stethoscopes are simple devices, and they are typically designed
for observing body sounds for short periods. In order to make them suitable
for long-term monitoring, a detection algorithm for the sound type of interest is
required. Various researchers have studied automatic and semiautomatic detection
of intestinal sounds. Some techniques such as fractal dimension [6], wavelet
decomposition [7], and neural networks [8] are used for short-term intestinal sounds
analysis. Other studies by Dimoulas et al. [9], Kim et al. [10], and Emoto et al. [11]
focused on more subtle change detections observed over longer durations.
For long-term monitoring of mobile users, another problem is the varying noise
that exists at every level and at every location of our daily lives. Two types of noise
require extra attention: environmental noise and body sounds. Some studies such as
Hadjileontiadis et al. [12, 13] and Dimoulas et al. [14] focused on signal denoising
and wavelet-based noise reduction. Jatupaiboon et al. [15] developed an electronic
stethoscope prototype with adaptive noise cancellation. The proposed prototype
acquires cardiac sounds and noises with two different microphones integrated in
the stethoscope. LMS algorithm for adaptive filtering is applied. Jiao et al. [16]
proposed a gradient adaptive step size LMS algorithm that uses two adaptive filters
to estimate gradients more accurately, for biomedical applications. Some researches
such as Ulusar [17] utilized classification techniques such as naive Bayesian and
minimum statistics and spectral subtraction for noise attenuation.
8 IoT and Edge Computing as a Tool for Bowel Activity Monitoring 135
Fig. 8.1 Bowel sounds (quiet period, noise, single burst, multiple burst) and their corresponding
power spectrum
The intestinal sounds (IS) refer to sounds produced within the small and large
intestines during digestion. They are typically characterized as a nonstationary burst
like signal and described as gurgling, rumbling, growling, and high pitched. These
sounds are occasionally contaminated by other artifacts such as heartbeat, move-
ment, breathing, and environmental noise [18]. Sound level varies and correlates
with the intensity and the location of the bowel movement. Bowel sound (BS) is
typically classified into two types: single burst (SB) and multiple bursts (MS) [18,
19]. Duration of single BS varies between 0.02 and 0.1 s and creates a spiky sound.
A detailed work on abdominal sound classification can be found in [20]. Figure 8.1
depicts some of the observed intestinal sounds.
8.3 Methodology
Two electret microphones (POM-5246P-R) are used for bowel activity monitoring.
The low-amplitude signals of the microphones were amplified with operational
amplifier (LM358) and were digitized using a 10-bit ADC at a sampling frequency
of 2048 Hz. The bias voltage of the operational amplifier circuit was set to be 1.55 V.
The system’s operating voltage was 3.3 V. Figure 8.2 shows the sensing device
design.
136 U. D. Ulusar et al.
Figure 8.5 shows the one circuit diagram of operational amplification unit applied
on both channels.
Figure 8.6 shows the custom box which was designed to hold the microphones,
electronic parts, and the battery. In order to amplify body sounds, the bottom part has
a conic shape similar to a stethoscope chest piece. Microphone 1 is located inside
the stethoscope’s conic metal part in order to obtain intestinal sounds. Microphone
2 is set to the upper side of the bioacoustic sensor to obtain ambient noise. Covering
box, made by a 3D printer, is utilized to hold the electronic circuit and battery.
The sound intensity drops with distance, and microphone orientation plays an
important role on which sounds will be acquired. By utilizing two microphones,
we aimed to observe environmental sounds with one microphone and the body
138 U. D. Ulusar et al.
Fig. 8.7 90-s recordings from the internal and peripheral microphone
sounds with the other. The internal microphone (Mic.1) observes bowel sounds and
ambient noise. Peripheral microphone (Mic.2) observes only ambient noise. Figure
8.7 shows 90-s-long sample recording. The blue line indicates the signals observed
8 IoT and Edge Computing as a Tool for Bowel Activity Monitoring 139
by the peripheral microphone, and the red line shows the internal microphone. The
effect of the environmental noise is visible in both recordings, but the intensities
vary.
Figure 8.8 shows a closer look of the observed noise by the microphones. Single
bursts with high intensities are observed at 1.2 and 4.5 s.
The ZigBee module has a 500 MHz RISC-based processing unit and 292 kB
internal memory. Most of the signal processing is performed on the edge device to
minimize transferred data. Only signal chunks detected as activity (intestinal or not)
are transferred for further processing. Data is processed as 10-s chunks buffered in
the edge device’s memory. Initially, 10-Hz high-pass filter is applied to remove DC
offset, and signals are normalized (Fig. 8.9).
Hilbert transform-based envelope is used on both signals, and activity regions are
detected by using an empirically chosen threshold value (0.02). Very short threshold
crossings (<0.05 s) are discarded. Figure 8.10 shows the application of the Hilbert
envelope on both signals.
140 U. D. Ulusar et al.
Fig. 8.10 Hilbert envelope and application on signals from the internal and peripheral micro-
phone. The black solid line shows the temporal envelope obtained using Hilbert transform
The start and the end time of activities and the signal above thresholds is
transferred to the edge computing unit with more processing power and memory
for further processing. In our case, this was a personal computer. Figure 8.11 shows
the signals obtained from both microphones and the data transferred to the edge
computing unit. There were 19 signal chunks for the peripheral microphone and 53
signal chunks for the signal obtained from internal microphone.
8 IoT and Edge Computing as a Tool for Bowel Activity Monitoring 141
D
p (x, c) = p(c) p (xi |c)
i=1
The class with the maximum posterior probability is assigned to the frame.
During training, an expert marks the chunks into two groups, BS and non-BS.
The algorithm learns the probability distribution of each class p(c) and the class
dependent feature distribution p(xi |c).
Finally, signals marked as BS are sent for further analysis. The intensity,
frequency, and the duration of the signals are used for long-term bowel activity
monitoring and abnormality detection.
8.4 Conclusion
IoT-related healthcare systems are low-powered networked devices that can collect
and share biophysical data over time and provide intelligent decision for preventing
serious illnesses. With the help of clinical decision support software, they have the
power to transform healthcare and provide more precise and timely decisions.
This work introduces an edge computing and IoT-driven monitoring system for
bowel sounds. Signal chunks of intestinal activity is marked and transferred to
cloud-based analysis system. The advantages of edge computing are utilized by
avoiding the need for energy expensive raw data communication.
References
1. F.M. Al-Turjman, Towards smart ehealth in the ultra large-scale Internet of Things era, in
2016 23rd Iranian Conference on Biomedical Engineering and 2016 1st International Iranian
Conference on Biomedical Engineering (ICBME), (2016), pp. 102–105
2. U.D. Ulusar, F. Al-Turjman, G. Celik, An overview of Internet of things and wireless
communications, in 2017 International Conference on Computer Science and Engineering
(UBMK), (2017), pp. 506–509
3. Y. Tang, G. Cao, H. Li, K. Zhu, The design of electronic heart sound stethoscope based
on bluetooth, in 2010 4th International Conference on Bioinformatics and Biomedical
Engineering, (2010), pp. 1–4
4. M.M. Rach, H.M. Gomis, O.L. Granado, M.P. Malumbres, A.M. Campoy, J.J.S. Martín, On
the design of a bioacoustic sensor for the early detection of the red palm weevil. Sensors 13(2),
1706–1729 (2013)
5. D.D.K. Patil, R.K. Shastri, Design of wireless electronic stethoscope based on zigbee. Int. J.
Distrib. Parallel Syst. 3(1), 351–359 (2012)
6. L.J. Hadjileontiadis, I.T. Rekanos, Detection of explosive lung and bowel sounds by means of
fractal dimension. IEEE Signal Process. Lett. 10(10), 311–314 (2003)
7. L.J. Hadjileontiadis, C.N. Liatsos, C.C. Mavrogiannis, T.A. Rokkas, S.M. Panas, Enhancement
of bowel sounds by wavelet-based filtering. I.E.E.E. Trans. Biomed. Eng. 47(7), 876–886
(2000)
8. K.-S. Kim, J.-H. Seo, C.-G. Song, Non-invasive algorithm for bowel motility estimation using a
back-propagation neural network model of bowel sounds. Biomed. Eng. Online 10, 69 (2011)
9. C. Dimoulas, G. Kalliris, G. Papanikolaou, A. Kalampakas, Long-term signal detection, seg-
mentation and summarization using wavelets and fractal dimension: A bioacoustics application
in gastrointestinal-motility monitoring. Comput. Biol. Med. 37(4), 438–462 (2007)
10. K.S. Kim, J.H. Seo, S.H. Ryu, M.H. Kim, C.G. Song, Estimation algorithm of the bowel
motility based on regression analysis of the jitter and shimmer of bowel sounds. Comput.
Methods Prog. Biomed. 104(3), 426–434 (2011)
11. T. Emoto et al., ARMA-based spectral bandwidth for evaluation of bowel motility by the
analysis of bowel sounds. Physiol. Meas.34(8), 925 (2013)
12. L.J. Hadjileontiadis, Wavelet-based enhancement of lung and bowel sounds using fractal
dimension thresholding-part II: Application results. I.E.E.E. Trans. Biomed. Eng. 52(6), 1050–
1064 (2005)
13. L.J. Hadjileontiadis, Wavelet-based enhancement of lung and bowel sounds using fractal
dimension thresholding-part I: Methodology. I.E.E.E. Trans. Biomed. Eng. 52(6), 1143–1148
(2005)
14. C. Dimoulas, G. Kalliris, G. Papanikolaou, A. Kalampakas, Novel wavelet domain Wiener
filtering de-noising techniques: Application to bowel sounds captured by means of abdominal
surface vibrations. Biomed. Signal Process. Control 1(3), 177–218 (2006)
15. N. Jatupaiboon, S. Pan-ngum, P. Israsena, Electronic stethoscope prototype with adaptive noise
cancellation, in 2010 Eighth International Conference on ICT and Knowledge Engineering,
(2010), pp. 32–36
16. Y. Jiao, R.Y.P. Cheung, W.W.Y. Chow, and M.P.C. Mok, A novel gradient adaptive step size
LMS algorithm with dual adaptive filters, in 2013 35th Annual International Conference of the
IEEE Engineering in Medicine and Biology Society (EMBC), (2013), pp. 4803–4806
17. U.D. Ulusar, Recovery of gastrointestinal tract motility detection using Naive Bayesian and
minimum statistics. Comput. Biol. Med. 51, 223–228 (2014)
18. C. Dimoulas, G. Kalliris, G. Papanikolaou, V. Petridis, A. Kalampakas, Bowel-sound pattern
analysis using wavelets and neural networks with application to long-term, unsupervised,
gastrointestinal motility monitoring. Expert Syst. Appl. 34(1), 26–41 (2008)
144 U. D. Ulusar et al.
S. Kanimozhi Suguna
Department of CSE, School of Computing, SASTRA Deemed to be University,
Thanjavur, Tamil Nadu, India
S. Nanda Kumar ()
CII Institute of Logistics, Chennai, Tamil Nadu, India
The collected data are fed into a control tower that aggregates them every 5 s to
provide near-real-time information across the entire fleet. Deutsche Bahn has used
these data to improve risk management practices such as real-time rerouting and
optimization, considering all existing network traffic through nodes [45]. Whirlpool
is another example of using the IoT for internal supply chain optimization in routing
work and locating misplaced inventory. Instead of using bar codes or a similar
solution, Whirlpool used radio-frequency identification (RFID) tags and readers
across a manufacturing plant to give managers and operators real-time access to
information for inbound logistics to the paint line [45].
According to the Gartner Press Release [21], the following technologies will
be extensively used in the coming years to become digital businesses. These
technologies are Bioacoustic Sensing; Digital Security; Smart Workspace; Con-
nected Home; 3D Bioprinting Systems; Affective Computing; Speech-to-Speech
Translation; Internet of Things; Cryptocurrencies; Wearable User Interfaces; Con-
sumer 3D Printing; Machine-to-Machine Communication Services; Mobile Health
Monitoring; Enterprise 3D Printing; 3D Scanners; Consumer Telematics. The IoT
can be used by the various partners of the supply chain to monitor the supply
chain execution process in real time and improve the efficiency and effectiveness
of supply chain [53]. Recent developments in the IoT have made it possible to
achieve high visibility in the supply chain. The location of arbitrary individual
things can be determined at any point in time by all appropriate supply chain
partners [22]. For example, the IoT benefits the food and agricultural product
supply chain by improving the visualization and traceability of agriculture products
and ensuring people’s food safety [66]. Industrial deployment of the IoT provides
development of an ideal platform for decentralized management of warehouses
and collaborative warehouse order fulfillment with RFID, ambient intelligence, and
multi-agent system [56].
Ravi and Arora [3] describe organizations as constantly engaged in actions. If an
organization does not take actions, it cannot remain solvent, and if the organization
consistently exercises poor judgment in the actions that it takes, it is destined to
fail. Organizations do not have minds, and they do not make decisions, they are
institutions that empower individual agents to make decisions on their behalf, and
empower other individuals to carry out the actions associated with these decisions.
These decision making agents can be individuals or groups. The complex nature of
organizational decisions creates information overload for individual agents, causing
them to engage in a number of information processing shortcuts, which threaten
the quality of their decisions. The ability to access infrastructures, platforms, and
software services based on what is needed and paying for only what is used put
start-up enterprises and small and medium enterprises at an advantageous position
vis-à-vis the competition from larger enterprises and firms. This also gives them
an edge in the market and an almost an equal position with much larger firms and
enterprises. This provides a level playing field and can put pressure on enterprises
that have not yet realized the potential and impact that cloud computing can have
when developing or implementing their strategies. There can also be pressure when
the organizational culture is not open to emergent market opportunities and has a
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 147
Another approach to see the production network is through the SCOR display, which
was created by the Supply Chain Council (SCC) [69] to instruct, comprehend, and
oversee supply chains. It is a model to both characterize and measure the execution
of an association’s production network.
The SCOR display is sorted out around the five noteworthy administration forms
(see Fig. 9.1):
• Plan: Alignment of assets to request
• Make: Conversion or esteem included exercises inside a production network
operation
• Source: Buying or procuring materials or administrations
• Deliver: All client collaboration, from getting a request to conclusive conveyance
and establishment
• Return: All procedures that turn around material or administration streams from
the client in reverse through the store network
This gives a wide definition to the inventory network, which features its
significance to the association and how it makes measurements to gauge execution.
The objective of the present store network is to accomplish combination through the
joint effort to accomplish visibility downstream toward the client and upstream to
providers. As it were, a large number of the present organizations have possessed
the capacity to “substitute data for stock” to accomplish efficiencies. The times
of having “islands of robotization,” which may enhance association’s inventory
network at the cost of another person’s (for instance, provider), are finished.
The ideas of collaboration and basic reasoning helped by innovation empower
associations to work with different capacities inside and with different individ-
uals from their store network, including clients, providers, and accomplices, to
150 S. Kanimozhi Suguna and S. Nanda Kumar
Production network and logistics capacities were seen fundamentally as cost focuses
to be controlled. It is just in the previous 20 years or with the goal that it has turned
out to be evident that it can be utilized for an upper hand also.
To achieve this, an association ought to set up aggressive needs that their
production network must need to fulfill inner and outer clients. They should then
connect the chosen aggressive needs to their store network and coordination forms.
Krajewski et al. [67] recommend breaking an association’s aggressive needs into
cost, quality, time, and adaptability capacity gatherings:
152 S. Kanimozhi Suguna and S. Nanda Kumar
Figure 9.3 demonstrates the different IoT applications in the supply chain. Supply
chain advancement is basic to practical upper hands for firms. Supply chain
innovation has been defined as “a change (incremental or radical) within a supply
chain network, supply chain technology, or supply chain process (or a combination
of these) that can take place in a company function, within a company, in an
industry or in a supply chain in order to enhance new value creation for the
stakeholder [2]. Production network information will progressively be created by
a system of sensors, RFID labels, meters, actuators, GPSs, and different gadgets
and frameworks [11]. IBM [30] recommends the smarter supply chain of the
future has the instrumentation, interconnectedness, and knowledge to anticipate,
if not forestalled, disturbances before they happen. From generation floor and
warehousing to the conveyance and store racking, the IoT empowers production
network accomplices to make brilliant inventory network by giving continuous
information and business knowledge for all accomplices in the inventory network.
Firms will put resources into the IoT to build deceivability of materials stream,
decrease the loss of materials, and lower circulation costs. This is delineated in
Fig. 9.4.
As the IoT enters into center business forms and an expanding number of
firms put resources into green store network, utilizing the IoT for keen production
network will turn into a subject of awesome enthusiasm for the chiefs [40]. The
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 153
IoT for keen store network will produce an incentive in stock administration,
preventive support, and transportation and permit multipath interchanges with the
accomplices and clients, rich information examination, and opportune reactions to
sudden occasions. Notwithstanding, the estimation of the IoT can be acknowledged
just when directors completely comprehend what information the IoT gadgets
gather, how the information ought to be handled, and when they should settle on the
choices on store network issues. It is likewise essential to break down specialized
and administrative security dangers of the IoT [35].
General engineering will have a noteworthy bearing on the field itself and should
be researched [24]. For effective IoT-based production network administration,
fundamental IoT engineering should be built up and refreshed persistently to
commission and decommission different IoT resources. Reaidy et al. [56] propose
an IOT foundation for synergistic distribution center request satisfaction in light of
RFID, surrounding insight, and multi-operator framework. It comprises a physical
gadgets layer, a middleware surrounding the stage, a multi-operator framework, and
an undertaking asset arranging. It incorporates a base up approach with choice help
systems, for example, self-association and transaction conventions.
For brilliant store network, the essential engineering incorporates four layers
[40]: (1) question layer, (2) correspondence layer, (3) application layer, and (4)
Fig. 9.4 Cost reduction and optimal distribution network design using IoT
information benefit layer. Each layer has center segments and basic functionalities
depicted underneath: object layer comprises physical protests, for example, gadgets,
machines, sensors, RFID labels, and peruser. The question layer is in charge of
detecting nature, recognizing and following articles, and gathering information.
The physical articles persuade scaled down to be more vitality productive, area
autonomous, and financially savvy [59]. RFID labels and sensors are frequently
installed in the machines and gadgets.
Application layer comprises an arrangement of issue particularly programming
devices that associate with clients, take care of issues, process information, and
offer arrangements with different applications. The application layer is in charge
of displaying information and data to the clients in an easy to use arrange. While
gadget-to-gadget applications do not really require information perception, more
human-focused IoT applications give representation to exhibit data to end clients
in a natural and straightforward route and to permit collaboration with the earth. It
is critical for IoT applications to be worked with knowledge, so the gadgets helps
in working with it. A standout among the essential results of the IoT is a gigantic
measure of information created from gadgets associated with the Internet [24]. At
the demand of the clients, a continuous examination of store network information is
led for basic leadership. The cloud information benefit framework ought to have the
capacity to powerfully organize demands and arrangement assets to such an extent
that basic solicitations are served continuously [24]. Information mining innovations
are utilized to find learning covered up in the ocean of information [10]. Not at
all like the other three layers which are regularly possessed by an individual firm,
the information benefit layer is frequently claimed and overseen by open cloud
specialist organizations.
The IoT encourages the improvement of heaps of industry-situated and client-
particular IoT applications [40]. Although gadgets and systems give the physical
availability, IoT applications empower the gadget-to-gadget and human-to-gadget
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 155
where people with detecting and processing gadgets on the whole offer information
and concentrate data to gauge and guide marvels of regular intrigue [20]. The
sensor arrangement is advancing toward a setting mindful of self-sorting out, self-
repairing, and self-sufficient system. Innovation for data sharing and coordinated
effort is moving to client arranged, conveyed insight framework, things-to-things
joint effort situations, and things-to-people cooperation. For more secure data
sharing and coordinated effort between gadgets, the rising innovation will reinforce
interoperability, network, protection, and knowledge highlights of the IoT. The fate
of enormous information and information investigation incorporates setting mindful
information handling, subjective preparing, and enhancement [41].
Supply chain IoT gadgets create a huge measure of information that should
be gathered and prepared progressively to give data on status, area, usefulness,
and condition of items and administrations. Since a lot of sensor information
continuously will build workloads of gadgets at an exponential rate, it is not practical
to process all the gushing information accessible to those gadgets. Setting mindful
information preparing empowers sensors and gadgets to utilize setting particular
data, for example, area, temperature, and the accessibility of a specific gadget to
choose what kinds of information to gather and decipher to give pertinent data to
basic leadership. For instance, setting mindful information handling can convey
significant data to a truck driver by knowing the truck’s flow area (e.g., on an
interstate, a congested inward city zone, or at a dangerous assembling plant). By
incorporating the human perception process into IoT applications, an intellectual
information handling application figures out how to detect things, foresee future
occasions, and frame conclusions in view of confirmation under various conditions.
For instance, intellectual information preparing utilizes computerized reasoning
procedures to comprehend the warehousing condition, forms ecological informa-
tion, for example, chemicals and contamination for a director, and uses an input
from the administrator to learn further. The enhancement of information preparing
is basic to an opportune procedure of the persistent stream of gigantic measures of
information. Mechanical advances in streamlined information handling help settle
on opportune choices in huge information applications in the keen store network.
Figure 9.6 underneath demonstrates the different IoT organization structures amid
the distinctive eras, past, present, and future.
IoT innovation stage is a mechanical establishment and improvement methodol-
ogy with which each organization builds up their IoT applications for new esteem
age and the upper hand. These stages are produced for the most part by IT firms
and utilized by non-IT firms. IT firms set IoT models and build up an assortment of
foundational advances for modern machines, ecological sensors, home apparatuses,
and medicinal services gear. The decision of the IoT innovation stages by non-IT
firms ought to be made regarding the general business and innovation methodology.
As of now, there are four generally embraced innovation stages: exclusive IoT
innovation, open development IoT innovation, industry union IoT innovation, and
IoT as a Service (TaaS).
Exclusive IoT innovation has been created by many firms to procure an early-
mover advantage, to end up plainly an innovation pioneer, and to be marked as a
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 157
trailblazer in their industry. For instance, Apple has its own licensed exclusive IoT
innovation for in-home gadget correspondence, HomeKit. Exclusive IoT innovation
is broadly produced for checking and control applications.
Open development IoT innovation is frequently utilized by a firm whose
advances are constrained in making esteem included IoT applications. Open inno-
vation has been viewed as applicable basically to “high-innovation” enterprises and
has gotten more prominent consideration in light of the verbal confrontation about
globalization and the potential for the R&D capacity to be outsourced [14]. Open-
source IoT stages are favored and are more alluring contrasted and the exclusive
IoT innovation stages because of the accompanying reasons. To begin with, the
utilization of the open source is relied upon to empower the speedier combination
of new IoT arrangements over the application spaces [49]. Second, the utilization of
the open source has been accounted for to accelerate the appropriation of a product
innovation in a base up form. At last, when seen from the social surplus point of
view, the industry in view of the open-source stages has been found to give bigger
aggregate welfare to the business structures [16].
Open advancement is known to quicken development process and make coop-
erative energy between endeavors of various innovation abilities. The main organi-
zation starts the open advancement and makes an open development coordination
with other littler organizations which contribute their mechanical know-how to a
particular issue. For instance, the Kimberly-Clark Digital Innovation Lab (D’Lab)
conveyed open-source development to the buyer bundle merchandise class with the
second yearly KChallenge start-up rivalry at the 2015 Consumer Electronics Show
(CES) in Las Vegas [36]. The triumphant new businesses will get the chance to
pilot a task with one of Kimberly-Clark’s worldwide brands, for example, Depends,
Scott, and Kotex, to discover better approaches to interface with their customers.
Kimberly-Clark fills in as an orchestrator and a facilitator of advancement, and
taking an interest in new company assumes a basic part in propelling development
for different brands of Kimberly-Clark.
158 S. Kanimozhi Suguna and S. Nanda Kumar
Based on the technology trends, this section discusses three capabilities of IoT-based
supply chain applications: (1) monitoring and control, (2) information sharing and
collaboration, and (3) big data and IoT data analytics.
As per the Digital Agenda for Europe [17], observing and controlling allude to
the control of any framework, gadget, or system through computerized techniques
oversaw by a control unit with or without the capacity to show data. IoT-empowered
checking and control frameworks are picking up fame in the natural and assembling
areas, where organizations have worked fundamental applications, for example,
constant perception with sensors and cautioning about machines and gadgets [7,
39, 60]. The IoT-empowered observing and controlling innovations are basic in
numerous business forms identified with inventory network exercises, for example,
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 159
Assets that put resources into store network connections empower data sharing and
coordinated effort between inventory network accomplices [51, 54]. To accomplish
SCM destinations effectively, firms need to lead a progression of SCM exercises.
160 S. Kanimozhi Suguna and S. Nanda Kumar
Inventory network accomplices must see exact, current, and solid data from different
firms with a specific end goal to encourage data sharing and joint effort [52].
To boost SCM’s maximum capacity, inventory network accomplices require data
sharing and coordinated effort. Data sharing and coordinated effort allude to the
electronic development of data and reason-driven cooperation between individuals,
among individuals and things, and between things.
Data sharing and collaboration applications improve the basic leadership of the
channel accomplices through better deceivability of material stream in the produc-
tion network. Between authoritative information sharing turns out to be ongoing
by means of remote interchanges. As production network accomplice’s increment
data sharing and cooperation exercises, they lessen between firm vulnerability and
reinforce vital relationship [37, 42].
Remote sensors are generally utilized for data sharing and joint effort among
items and human leaders. Undertakings can enhance their abilities by utilizing
the IoT-related data sharing systems inside and crosswise over enterprises [43].
Between gadget data sharing and coordinated effort, there is an advantage from
the institutionalization of correspondence conventions and gadgets. Checking and
control segments are frequently implanted into the data sharing and cooperation
gadgets and identify a predefined occasion which triggers data sharing and joint
effort among individuals and things. Late research demonstrates that there exists
more potential use of IoT in data serious divisions, for example, medicinal services
benefits, because of data sharing capacity of the IoT [64]. Data sharing enables
gadgets and individuals to work together in the meantime, with a similar data
gathered or recovered from numerous sources. Store network accomplices can share
data created by IoT gadgets to track between authoritative calculated streams of
materials and items and team up their conveyance exercises in view of the level of
stock and the rate of offers. The data sharing and joint effort will be more gainful
for worldwide production network accomplices where the vulnerability of lead time
is more prominent than for the residential store network accomplices.
One key innovation of the IoT for data sharing and joint effort is fast system inno-
vation [62]. The capability of the IoT for production network will be acknowledged
by means of IoT institutionalization with which gadgets from various makers share
data consistently among inventory network accomplices. Measures are required for
bidirectional correspondence and data trade among things, their condition, and their
advanced partners that have an enthusiasm for checking, controlling, or helping the
things [5].
A cloud computing architecture will be highly suited in such data sharing
environments, integrating the full extent of capabilities that cloud providers offer
with internal resources to provide a comprehensive technical support and delivery
solution. Cloud computing promises to enhance and optimize the technical capabil-
ities of enterprises. Resource sharing in a multitenant infrastructure enables clients
of cloud computing services who independently may not be able to make the invest-
ment for the newest technology solutions or to create the redundancy necessary to
provide 100% availability. Resource constraints have caused enterprises to attempt
to do more with fewer resources or to plan for redundancy only for the most critical
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 161
Big data and data analytics allude to the way toward gathering, arranging, and
breaking down a huge measure of information to find helpful examples and learning
[19, 41] defines big data as data that cannot be handled and processed in a
straightforward manner. A big dataset won’t fit in memory unlike a spreadsheet
data that fits easily in memory. Big data is not easy to check whether it is clean.
Computations will take a long time. In big data, new data will be constantly
streaming in, and the processing system must make decisions as to which part of
the stream must be captured. The dataset may consist of images, natural language
162 S. Kanimozhi Suguna and S. Nanda Kumar
text, and other heterogeneous data. Also, big data dataset will probably be so large
as to not fit on a single hard drive; and hence, will be stored on several different
disks across a network and will be processed on a number of cores. Queries will
have to be distributed and written to work across a network.
Also, frequent patterns generated over data streams also become large, which
means spending a lot of time mining the patterns, and thereby it can violate
one of the requirements for the data stream mining, immediate processing. In
order to solve the problem, closed frequent pattern (CFP) and maximal frequent
pattern (MFP) notations [55]. Huge information presents both enormous open
doors and huge difficulties to our general public. The open doors incorporate
better-educated business choices, more proficient store network administration and
asset assignment, more compelling focusing of items and notices, and speedier
turnaround of logical revelations [38]. IoT gadgets and machines with installed
sensors and actuators produce a colossal measure of information to be handled
by business insight and examination apparatuses for basic leadership. Various
examinations concentrate on utilizing or creating compelling techniques to change
over the information produced or caught by the IoT into learning helpful for basic
leadership and critical thinking [8, 61]. Imaginative IoT applications are utilizing
consistent extensive scale detecting, information investigation, data portrayal, and
distributed computing [24].
Information created by the various accomplices’ machines and gadgets are
nourished into a cloud data center and converged into a production network-
wide database for reproduction, information investigation, and advancement. Enor-
mous information and information investigation applications improve the prescient
upkeep of the Armada, request gauge, conveyance lead time, and client administra-
tions. The enormous information produced by IoT has diverse attributes contrasted
and general huge information due to the distinctive sorts of information gathered, of
which the most traditional qualities incorporate heterogeneity, assortment, unstruc-
tured element clamor, and high repetition [13].
The IoT makes enormous open doors for the economy and people. The IoT has no
limits and will affect each business and individual. With new gadgets associating
into organization’s frameworks, new kinds of information will be open. No less
than three of the real advantages are correspondence, control, and cost investment
funds which prompt changed business forms, enhanced client maintenance, new
plans of action, and an expansion in the personal satisfaction. An investigation done
by McKinsey Global Institute (2015) predicts that the IoT has a potential monetary
effect of $3.9 trillion to $11.1 trillion every year in 2025 – the estimation of this
effect would be proportionate to roughly 11% of the world economy in 2025.
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 163
As the Internet of Things (IoT) opens up new business openings, firms are
attempting to comprehend the effects of the IoT transformation on their inventory
network. Regardless of the enormous enthusiasm by network administrators to use
the IoT, there is yet a lack of arrangements with the innovative part of the inventory
network.
Cloud computing represents an evolution in how enterprises acquire and use tech-
nology from external cloud service providers and interact with and support internal
and external users and customers. Because of its evolutionary nature, clear guidance
is needed and a direction for executive management or those responsible for
determining as to how the cloud will be used and how it fits within the organizational
structure and processes. Further evolution will let to better understanding, bringing
more specific guidance addressing the legal, operational, and technical types of risk.
Such guidance will help reduce the need to plan for building trust to achieve value
and benefit at each step in the acquisition and integration of cloud solutions.
Similarly, as with any troublesome advancement, the IoT exhibits different
difficulties for firms. This segment recognizes and talks about difficulties in IoT
application advancement for shrewd inventory network. For instance, Gubbi et al.
[24] propose that because of the blast of the information created by IoT machines,
server farms will confront challenges in security, nature of administration, informa-
tion investigation, stockpiling administration, server innovations, and server farm
organizing. Khan et al. [34] recognize challenges from specialized points of view:
naming and personality administration, interoperability and institutionalization, data
protection, objects wellbeing and security, information privacy and encryption,
organize security, range, and greening of IoT. Many investigations transcendently
call attention to the security test of the IoT [18, 32, 57]. This segment talks about five
specialized and administrative difficulties: information quality, information security,
speed of innovative change, store network process upgrade, and venture avocation
[40]. The talk concentrates on how these difficulties may have impacts on the
advancement and reception of production network IoT applications.
of the information will be low, and the reception of IoT advances can be blocked. At
the point, when information gathering cost is high and the nature of the information
is low, the estimation of the IoT diminishes quickly. In different circumstances,
the information required for basic leadership is not accessible or not convenient.
On the off chance that accomplices do not give opportune information on the sum
and timing of a buy, providers would not have the capacity to build up a powerful
generation design [12]. For the observing and control IoT applications, information
quality is basic to the estimation of and trust in the IoT. For firms embracing the
IoT applications for their inventory network, machines may not work appropriately,
information gathered may not be dependable, and basic choices might be made
by clients or machines in light of erroneous information. Production network
accomplices must see information from different firms to be precise, current, and
solid with a specific end goal to encourage data sharing and joint effort [52].
Directors would not utilize enormous information and information examination IoT
applications in the event that they do not have a trust in the information quality and
might not have any desire to impart their information to their accomplices.
As IoT is based on the Internet, security issues of the Internet will likewise happen
in IoT [32]. Once the IoT information is spilled or hindered in any piece of the
IoT framework, the security of the whole IoT framework will be under threat.
As the quantity of associated gadgets builds up, the quantity of the conclusion
to end association focuses increments exponentially, and the quantity of potential
security gaps increments in extent. Securing the IoT is an unpredictable and
troublesome assignment. The number of assault vectors accessible to pernicious
assailants may end up plainly amazing, as a worldwide network (“get to anybody”)
and openness (“get to in any case, whenever”) are key principles of the IoT [57].
These security openings to singular observing and control IoT applications open
the entryways for security breaks, for example, dissent of-benefit (DOS) assaults
and Trojan steed assaults from programmers and other digital hoodlums. For store
network cooperation, the information gathered from the IoT applications should be
safely ensured to keep the privacy of inventory network accomplices’ information.
A current report by HP uncovers that 70% of the most regularly utilized IoT
gadgets contain genuine vulnerabilities with secret word and the client have to
get authorizations for it [28]. Filtering of ten most famous IoT gadgets uncovers,
by and large, every gadget contained 25 powerless gaps or dangers of trading off
the IoT security. The IoT gadgets normally do not utilize information encryption
procedures. Feeble security will jeopardize the client protection from the reception
of huge information and information examination of IoT applications for brilliant
store network. Security test might be reduced via preparing application designers
to join security arrangements, for example, interruption anticipation frameworks,
encryptions, and firewalls into items, and urging clients to use the IoT security
including those that are consolidated into their gadgets.
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 165
The speed of mechanical change alludes to the rate of innovative change in the
IoT. Development is quickening at a breathtaking rate in numerous IoT innovation
wildernesses. The union of advancements in objects, systems, information stock-
piling, and applications are encouraging the speed of the IoT development. The
expansion in the speed of the innovative change prompts a more prominent level of
natural vulnerability [6, 25]. The quick mechanical improvement in the IoT is fixing
to the estimation of the market of $7.1 trillion by 2020, up from $1.9 trillion of every
2013, as the development of market request drives the innovative advances in the
IoT. Remote sensor systems (WSNs) have been perceived as a promising innovation
for the IoT, and mechanical advances in equipment have made hearty and practical
gadgets in detecting applications [58]. Brilliant store network is being produced with
propels in sensor advances and scaled down installed computational gadgets. Since
innovative changes cannot be managed by an individual firm, firms need to fabricate
a nearby business relationship to mutually grow new advancements and items [47].
Mendelson and Pillai [46] demonstrate that as the business condition turns out to be
more questionable, firms need to refresh and overhaul existing inventory network
through coordinated effort methodology all the more regularly. While different joint
effort IoT applications increment the general store network efficiency, it will be
harder to anticipate which IoT innovation will turn into a troublesome development
in production network administration. The topic of what firms ought, to limit the
danger of removal expedited by the IoT’s problematic advancement stays to be
replied.
The IoT is an empowering influence of store network process update and in addition
a problematic advancement. In the meantime, this new advancement brought out
new difficulties for the outline of progressively more intricate supply chains.
Changing shopper practices and the rise of transformative advances, for example,
the Industry 4.0, counterfeit consciousness, huge information examination, and
the Internet of Things, are driving a progression of developments with significant
ramifications for the endeavors in their plan of supply chains [27]. While the
utilization of the IoT in-store network is required to enhance responsibility in
associations, it is indistinct how such desires will really be met in light of the fact
that new advances address as well as create new responsibility requests and directors
may encounter challenges in fulfilling these accountabilities [9].
Supply chain process overhaul endeavors mean to enhance arrange satisfaction,
stock administration, warehousing, conveyance administration, and client relation-
ship administration. Inventory network IoT’s potential esteems talked about in the
calculated structure can be completely acknowledged by coordinating it with store
166 S. Kanimozhi Suguna and S. Nanda Kumar
network process overhaul. Devaraj and Kohli [15] find that IT capital venture
joined with business process reengineering enhanced gainfulness in the social
insurance industry. While checking and control IoT applications may effectively
meet operational proficiency, they may at present be ineffectively coordinated with
joint effort concentrated production network procedures, for example, anticipating
and arrange satisfaction. Subsequently, firms may make storehouses of checking
and control IoT applications without completely understanding the advantages
of community IoT applications. New supply process overhaul may require work
investigation, work disposals, work creation, and employment preparation because
of the end of existing procedure exercises and production of new process exercises.
A change administration program for influenced inventory network experts should
be set up before the selection of the IoT.
While the IoT is generally new, speculation openings flourish alongside the
advances of different IoT innovations. Imaginative firms are relied upon to use
different capacities of the IoT advances. While actualizing an IoT domain requires
enormous speculations. The venture choice where things are not open and inter-
operable makes it troublesome for organizations to embrace this innovation [1].
To understand the potential advantages, firms require a fitting measure with a
specific end goal to legitimately survey the dangers and prizes of the innovation
[41]. Numerous IoT ventures have vague issue definitions and are utilizing rising
advances, along these lines, causing a higher danger of task disappointment and
higher irreversibility of speculations than customary innovation ventures. When
all that is said is done, interests in huge information and information examination
IoT applications are more unpredictable and less secure than interests in the other
two IoT application classifications. The standard measure firms commonly use to
esteem ventures, net present value (NPV) and return on investment (ROI), frequently
may not catch genuine estimation of the IoT extension because of the way these
standard measures disregard adaptability in speculation, for example, reversibility
and versatility in the assessment time frame. NPV tends to dismiss a venture’s
worth when there is the high vulnerability of undertaking achievement and is not
reasonable for high hazard ventures, for example, ventures for huge information
and information investigation. So as to esteem high hazard IoT ventures, the genuine
choice approach might be more proper [41].
The benefits of cloud acquisition based on a full understanding of the cost of
cloud compared with other technology platform business solutions are evaluated.
The full cost of provisioning technology infrastructures and services is much larger
than the cost of facilities, hardware, and software and the annual cost of human
resources. The full cost of acquiring, deploying, and maintaining cloud infras-
tructures, platforms, or software services is more than what is represented within
the cloud contract and SLA. Enterprises must also make a substantial investment
9 Application of Cloud Computing and Internet of Things to Improve Supply. . . 167
9.7 Conclusion
Because the IoT is such a recent development, there remains a paucity of studies on
the supply chain aspects of the IoT. This makes it very challenging for enterprises
to make informed decisions in regard to IoT adoption/implementation. Many enter-
prises invest in a variety of IoT applications for their supply chain innovation. The
IoT became a reality beyond the hype. However, converting the IoT investment into
real performance and profit improvement is challenging. Our chapter contributes
to the literature in this emerging field by presenting a foundational knowledge for
building a smart supply chain.
This chapter presented a conceptual framework of the Internet of Things (IoT)
for smart supply chain management and identified three capabilities of the IoT:
monitoring and control, information sharing and collaboration, and big data and
data analytics. This chapter also examined how the IoT applications support the
supply chain management of Fortune 200 companies. This chapter presents core
and emerging IoT technologies and technology platforms for a smart supply
chain. Finally, this chapter discussed five technical and managerial challenges in
implementing IoT applications for smart supply chain: data quality, data security,
the velocity of technological change, supply chain process redesign, and investment
justification.
168 S. Kanimozhi Suguna and S. Nanda Kumar
References
10.1 Introduction
Ubiquitous connectivity is the need of the hour with the ever-increasing volume of
devices getting access to the Internet. Several key technologies drive the prolifera-
tion of IoT solutions across multifaceted application domains that are extensively
discussed, researched, and adopted. The key challenge in IoT is processing the data
from devices and performing meaningful interpretation to derive insights and take
decisions.
Cloud computing has made possible the widespread adoption and proliferation
of IoT. It has been the backbone of IoT deployments and applications which has
helped in achieving primary network objectives like storage, communication, and
computation. It also offers services that can be tailored according to user and
application requirements. Cloud computing consists of a pool of resources, typically
consisting of services like networking, storage, and computing, which are organized
to cater the needs of multiple end users based on a multitenant model. The access
to these resources is via standard mechanisms and is usually available across a
network. The varieties of service offered by the cloud computing paradigm include
SaaS (Software-as-a-Service) in which applications are offered as service, IaaS
(Infrastructure-as-a-Service) which offers computing resources as a service, and
finally PaaS (Platform-as-a-Service) which provides software and hardware tools for
development. The widely adopted deployment models of cloud computing include
public and private clouds in which organizations deploy cloud services to any end
user or within their own cloud computing platform, respectively. Cloud computing
is capable of providing the following value additions – elasticity, ubiquity, reduced
management effort, pay-as-you-use, and convenience which has been a driving force
for the creation of a rapidly growing industry across the globe which has its worth
in billions [9].
Regardless of all these benefits, there are few areas in which cloud computing
does fall short of meeting the application requirements, which include increased
network latency, jitter, lack of ability to access local contextual information, and
lack of support for mobility of users. For an application that is delay sensitive,
these requirements must be met. Due to these reasons, there has been an evolution
of new computing paradigms such as fog computing, mobile cloud computing,
and mobile edge computing, among others [2, 7]. Edge computing leverages the
resources available on the local edge to meet the specific application requirements
mentioned. Edge computing can bring the services offered by cloud computing to
the network edge and the essential functionalities like communication, storage, and
computation closer to the devices and users.
The primary objective of these computing technologies is to bring about the capa-
bilities of cloud computing closer to the user or near the network edge. Normally
most edge computing paradigms follow the structure as shown in Fig. 10.1. The
edge layer devices as specified can be microdata centers, gateways, and dedicated
Cloud Layer
Cloud Infrastructure
Edge Layer
Device Layer
Fog Computing
Table 10.1 gives an overall idea of how the edge paradigms differ from the tradi-
tional cloud structure. This comparison is necessary to understand the challenges
that arise when we need to implement solutions in edge computing. It is evident that
all the edge paradigms have the common goal of bringing the capabilities of cloud
computing closer to the edge. They all provide in some way the multitenant vir-
tualization infrastructure which helps in location-specific provisioning capabilities
and access to nearby computational resources when required. We have taken into
account MEC, MCC, and fog computing for our comparison [1, 5]. As mentioned
earlier all of them have own similarities among them and uniqueness as well. They
provide a similar set of benefits that are majorly due to the closeness of the edge
data centers. The benefits are as follows:
• Low and predictable jitter and network latency
• Location-based context-aware information
• Support for scalability
• High availability of services
Figure 10.3 summarizes the similarities and differences between the edge
paradigms. The predominant similarity is that all these techniques basically can
provide support for mobility. They have consideration for device mobility and
have specific management mechanisms that are located at the application level
itself or use virtual instances of devices to accommodate for the same. The next
similarity is the overall architecture which has support to enable the edge paradigms
to behave as an extension of cloud infrastructure. The network elements act in
a distributed and decentralized way capable of provisioning services and taking
decisions autonomously. They can also cooperate with each other to reduce the
dependency on the central cloud setup. The different edge data centers can exchange
information among them and coexist to form a federated infrastructure.
Moving on to the differences, even though the edge paradigms have a common
objective, they have few basic differences in the way they fulfill the objective. MEC
deployments primarily focus on realization of 5G, whereas fog nodes can provide
services to other applications which have their own servers, gateways, access points,
and so on. In case of MCC, it is highly distributed in nature, and device instances
can perform their own service provisioning. This difference in management and
deployment of the data centers determines who will be service providers. Another
difference is that the applications supported are determined by the choice of service
providers. MEC provides support for operators to work closely with other third-
party service providers, which enables for thorough testing and possibility for
customized integration. This applies true even for fog computing also. The final
difference is that MCC paradigm provides specific services that are not related to
virtualization but enables support for execution mechanisms that are distributed in
nature. This is highly beneficial for devices that have severe resource constraints.
176 Bragadeesh S. A. and U. Arumugam
Although there are differences, still the security and privacy techniques used for
one paradigm can be easily adopted for the other.
The key attributes whenever we are trying to implement security solutions are
authentication, integrity, confidentiality, trust, access control, data security, and
privacy [1, 11]. The various components of providing end-to-end security and
privacy are as follows.
Fig. 10.4 Summary of security and privacy requirements for edge computing
10 A Conceptual Framework for Security and Privacy in Edge Computing 177
This is required since the edge devices are distributed across the network and the
cost of maintenance is high. However, a new network concept called software-
defined networking (SDN) can come to the rescue. The benefits of SDN include but
not limited to reducing the burden of management and implementation, improving
the scalability of the network, decreasing the cost of maintenance, providing
prioritization and isolation of network traffic, and enhancing the access control of
network resources, cooperation, and network sharing.
Like cloud computing, the data storage in edge platform is also outsourced. It is
tough to ensure data integrity since there are a lot of possibilities for data loss and
data modifications. If the data present in one device is compromised, the attackers
could easily abuse the data stored to fulfill their own needs. Hence, provision for
data storage auditing should be done. Third-party auditing services are available,
which are provided by the infrastructure providers, and the application users should
be aware of the auditing policies. It is possible to use encryption techniques to
ensure integrity, verifiability, and confidentiality of the data to check for untrusted
network entities and allow the user to check the stored data. Since all data need
not be essentially present on all the available storage resources, reduced latency and
support for dynamic access are required.
The data computation performed at the edge servers and devices should be secured
and verifiable. The security of computations can be ensured by using data encryption
techniques, which prevents data visibility to any hackers/attackers. Since the
microdata centers have the provisions to off-load some of the computations to
other data centers, a mechanism to verify the computation results and establish trust
between the two entities becomes a necessity. The computational accuracy should
be verifiable by the user. Use of data encryption for the data that is being forwarded
to the edge data centers from end devices and from one data center to another
helps in ensuring data integrity and protection. One of the popular mechanisms to
verify the data security is using data search technique in which a keyword-based
search is performed in the encrypted data repositories. This enables users to securely
search for user data amidst encrypted data. This helps in maintaining the secrecy of
encryption.
Since the end devices are typically less powerful and have limited access to the
surroundings, they can be easily tampered with. Any hacker may try to take control
178 Bragadeesh S. A. and U. Arumugam
of a device and make it a rogue node to retrieve essential network management data.
They can hinder the normal behavior by corrupting the device data and increase the
frequency of data access by sending fake information. They can also use the end
device to propagate false data across the network to potentially create discrepancies
and disrupt the normal network behavior. Hence, proper efforts should be made for
ensuring the security of the end device.
Intrusion detection techniques help in identifying malicious data entries and detect
device anomalies. They can be used to carefully investigate and analyze the behavior
of devices in the network and provide methods to perform packet inspection,
which helps in early detection of denial-of-service attacks, integrity attacks, and
data flooding, among others. The primary challenge when implementing intrusion
detection for edge platform is to accommodate for scalability, mobility, and low-
latency requirements.
The threats can be grouped according to the component of the system they try to
attack [6]. The implementation of solutions to these attacks can be made possible by
a combination of security attributes such as authentication, confidentiality, integrity,
availability, and privacy.
Figure 10.5 represents the basic components of the proposed conceptual framework.
The framework consists of a hierarchical structure in which the communication and
exchange of data can be done using APIs. There are three levels: device manage-
ment, security services, and application and infrastructure. The proposed framework
aims to bring into account the necessary security and privacy requirements which are
specific to edge computing platform. The framework can be implemented in an edge
Cloud Infrastructure
Security Services
Access Control Authentication Trust Management Data Integrity
Policies Services Engine Verification
Device Management
System Data
Database Device Service
Management Orchestration
The primary functionalities of the device management layer are to manage the end
devices, create database, manage the services provide to the end device, and perform
data orchestration operations. The status of the end devices, access control, main-
taining integrity of the data stored, and providing the necessary data from service
providers are all taken care by the device management layer. Network function
virtualization (NFV) is one of the popular and emerging technologies which can
perform the tasks mentioned above. NFV in convergence with SDN can help in
realizing the full potential of edge computing platform. NFV provides virtualization
of infrastructure services, orchestration, and management functionalities.
The security services layer is the heart of the framework which performs the
critical tasks such as creating, updating, and maintaining the access control policies.
It also ensures authentication of devices, service providers, and the data that is
communicated. It also makes use of a trust management engine which is used
for accomplishing two objectives. It governs the privacy of the overall system and
generates rules for implementing intrusion detection mechanism. The data integrity
verification component performs the verification of data across the system and
computations done by other edge devices. Establishment of well-defined APIs and
integration of the modules with the application and device management layer help
in improving the overall performance of the framework and achieving the edge
computing specific requirements.
This layer is responsible for keeping track of the network-wide application interac-
tions and governs the authentication policies for accessing the network devices. The
service provisioning details of neighboring data centers and making the decision
of whether cloud support is required or not are carried out by this layer. Also, the
services offered by the infrastructure providers are monitored and provisioned for
access to the end devices. This layer is significant because edge platform caters to
multitenant application services, hence implementing security measures for it is an
essential task.
182 Bragadeesh S. A. and U. Arumugam
Edge computing has opened new arenas and created new opportunities to build
solutions that can augment existing cloud-based solutions. In order to explore the
full potential and bring about the widespread adoption of edge computing solutions,
one key area of focus is how to secure the system components and how to ensure the
privacy of user and organization data [6]. Figure 10.6 points out some of the major
challenges that need to be addressed [3, 4, 8, 10]. The challenges can be summarized
as below.
• The added security mechanism should not increase the computation overhead
and hamper the storage space required for other system operations.
• Edge computing mechanisms use cache management techniques which are prone
to side channel attacks; hence, care should be taken to prevent leakage of private
and sensitive data.
• For applications that stream data continuously for monitoring and management
purposes, hence due to increased number of devices, the volume data to be
processed is typically large. To detect the anomaly, packet filtering mechanisms
need to be implemented which might require additional memory and processing
power.
• It is difficult to manage user identity, scalability, monitoring, performance, data
security, and considering threats from insiders due to the large number of users
who share the resources and application in the edge computing environment.
10.6 Conclusion
In this work, we have defined the various edge paradigms and listed out their
similarities and differences. We have also made a comparison of edge computing
with cloud computing to drive in the importance of new implementation challenges
and approaches required for edge computing solutions. We have analyzed the
various aspects of security and privacy for edge computing and also brought into
light some of the possible threats. A conceptual framework for ensuring security
and privacy in edge computing has been discussed along with various components
involved in it. The reality to be considered here is that security and privacy solutions
for edge computing are still in their nascent stages, and, hence, we have analyzed
the possible challenges that might arise. This work is intended to be an eye-opener
for developers and technology enthusiasts who are involved in developing security
and privacy solutions for edge computing.
The future scope of this work is to develop a system which consists of a
knowledge base, rule-based inference engines, and a decision support toolbox that
could be added to an edge computing platform which helps in mitigating possible
security attacks and ensures the privacy of user data.
Acknowledgment The authors wish to express their sincere thanks to the Department of Science
& Technology, New Delhi, India (Project ID: SR/FST/ETI-371/2014), and express their sincere
thanks to the INSPIRE fellowship (DST/INSPIRE Fellowship/2016/IF160837) for their financial
support. The authors also thank SASTRA Deemed to be University, Thanjavur, India, for extending
the infrastructural support to carry out this work.
184 Bragadeesh S. A. and U. Arumugam
References
1. N. Abbas et al., Mobile edge computing: a survey. IEEE IOT J. 5(1), 450–465 (2018)
2. Y. Ai, M. Peng, K. Zhang, Edge cloud computing technologies for internet of things: a primer.
Digit. Commun. Netw. 4(2), 77–86 (2018)
3. F.A. Alaba et al., Internet of things security: a survey. J. Netw. Comput. Appl. 88(April), 10–28
(2017)
4. A. Alrawais et al., Fog computing for the internet of things: security and privacy issues. IEEE
Internet Comput. 21(2), 34–42 (2017)
5. A.V. Dastjerdi, R. Buyya, Fog computing: helping the internet of things realize its potential.
Computer 49(8), 112–116 (2016)
6. S. Khan, S. Parkinson, Y. Qin, Fog computing security: a review of current applications and
security solutions. J. Cloud Comput. 6(1), 19 (2017)
7. P.G. Lopez et al., Edge-centric computing: vision and challenges. ACM SIGCOMM Comput.
Commun. Rev. 45(5), 37–42 (2015)
8. B.A. Martin et al., OpenFog Security Requirements and Approaches, (November, 2017), pp.
1–6
9. R. Roman, J. Lopez, M. Mambo, Mobile edge computing, Fog et al.: a survey and analysis of
security threats and challenges. Futur. Gener. Comput. Syst. 78, 680–698 (2016)
10. J. Tan, R. Gandhi, P. Narasimhan, Challenges in Security and Privacy for Mobile Edge-Clouds
(Parallel Data Laboratory, Carnegie Mellon University, 2013). http://www.pdl.cmu.edu/PDL-
FTP/associated/CMU-PDL-13-113.pdf
11. S. Yi, Z. Qin, Q. Li, Security and privacy issues of fog computing: A survey, in
10th International Conference on Wireless Algorithms, Systems, and Applications, ed.
by K. Xu, H. Zhu (Eds), (Springer International Publishing, 2015), pp. 685–695.
https://doi.org/10.1007/978-3-319-21837-3_67
12. Y. Guan, J. Shao, G. Wei, M. Xie, Data security and privacy in fog computing. IEEE Netw. 99,
1–6 (2018)
Index
A C
Africa, mist computing challenges Cloud-based radio access network (CRAN),
assimilation and domestication, 125 33, 37
business blueprint, 126 Cloud computing
drivers supporting growth, 126–127 deployment clouds, 52
failure, faults and risk, 125–126 vs. edge computing, 176–178
funding opportunities, 126 mist and dew computing, 111
inclusiveness and governance, 125 service models, 52–53
skills gap and digital divide, 124–125 supply chain management
technical recommendations, 128–129 data quality, 164
technology transfer, 125 data security, 164–165
investment justification, 166–167
redesign process, 165–166
B technological change velocity, 165
Big data Cloud provider
IoT applications, 161–162 cloud service brokers, 11
mist and dew computing, 111 future improvement, 12
in smart cities, SDN, 92–93 Internet of Things
Bioacoustic sensor, 135–138 black box syndrome, 10
Blockchain technology, 115, 121, 123 communication models, 9
Bowel activity monitoring interconnectivity, 10
battery charger circuit, 137 notebook PC vs. tablet, 11
bioacoustic sensor, 135–137 privacy considerations and legal issues,
bowel sounds, 135 10
electronic stethoscopes, 134 smart objects, 9
high-pass filter and normalization, 139 volume, 9–10
intestinal activity detection, 141–142 pattern prediction technique, 8
intestinal sounds, 135 resource allocation
IoT-driven sensor device, 136 Amazon’s charging models, 3
LMS algorithm, 134 instance type, 4
operational amplifier circuit, 138 power consumption, 6–7
power spectrogram, 142 revenue, 5
regulator circuit scheme, 137 Cloud service brokers (CSBs), 11
signal processing steps, 137–141 Communications service providers (CSPs), 35
P computational methods, 96
Platform-as-a-Service (PaaS), 52 data collection and intermediate levels,
Private cloud, 52 95–96
Proposed system architecture, SDIoT, 93–94 data management and processing level,
application level, 100–101 96–97
computational methods, 96 HDFS system, 100
data collection and intermediate levels, message queue, 97–100
95–96 named data network, 101
data management and processing level, threshold limit lalue, 97
96–97 simulation results, 101–104
HDFS system, 100 Spark and GraphX, 101
message queue, 97–100 Supply chain management
named data network, 101 adaptability technique, 152
threshold limit lalue, 97 bolster exercises, 150
Public cloud, 52 cloud platform, IoT
data quality, 164
data security, 164–165
R investment justification, 166–167
Radio access network (RAN), 36 redesign process, 165–166
Radio applications cloud servers (RACS), 34 technological change velocity, 165
Radio network controller (RNC), 36 cost methodology, 152
Recurrent neural networks (RNNs), 43 CSCMP, 148
RedEdge model, 64 integrated value-added, 149–150
inventory network, 148
IoT applications
S big data and data analytics, 161–162
SDIoT, proposed system architecture, 93–94 cost reduction, 154
application level, 100–101 data sharing and collaboration, 159–161
computational methods, 96 innovation stage, 156
data collection and intermediate levels, logistics, 155
95–96 monitoring and control, 158–159
data management and processing level, network architecture, 157
96–97 open advancement, 157
HDFS system, 100 optimal distribution network, 154
message queue, 97–100 RFID labels, 153, 154
named data network, 101 streamline solutions, 162–163
threshold limit lalue, 97 Things as a Service, 158
Security-as-a-Service (SaaS), 52 leveraging effect, 151
Security mechanisms production network administration, 148
access control, 180 quality procedure, 152
data computation, 179 SCOR model, 149
data storage, 179 time system, 152
end devices, 179–180
intrusion detection, 180
network security, 178–179 T
Software-defined networking (SDN) Time division multiple access (TDMA), 37
big data in smart cities, 92–93
implementation and analysis, 101–104
proposed system architecture, SDIoT, W
93–94 Wireless medical sensor networks (WMSNs),
application level, 100–101 133