International Journal of Advanced Computer Engineering and Communication Technology (IJACECT)

_______________________________________________________________________________________________

A Survey of Resource Allocation Techniques in Cloud Computing

1
Kamini Bharti, 2Kamaljit Kaur
1,2
Department Of Computer Science & Engineering, Guru Nanak Dev University Amritsar, India
Email: [email protected], [email protected]

resources to competing requests based on pre-defined

Abstract----Nowadays Cloud Computing is considered as
one of the emerging areas in the various fields related to resource allocation policies. If the allocation is not
computer science in one way or another. Itdeals with on managed properly resource allocation starves services,
demand dynamic resource allocation for providing reliable this problem is solved by allowing the service providers
and guaranteed services to the consumer.It is based on to manage the resources for each individual module.
pay-as-you-use model to public. Resource allocation is Resource allocation is a part of resource management
performed with the objective of minimizing the costs and it is used to assign the available resources in an
related with it. The tasks of resource allocation are meeting economic way.
customer demands and application requirements. In this
paper, different resource allocation policies have been A. Cloud Architecture
surveyed based on the different parameters and the related
gaps have been discussed. The architecture for Cloud Computing can be divided
into three layers: Resource, Platform and Application.
Keywords: Cloud computing, resource allocation, The resource layer is the infrastructure layer which is
consumer context, virtualization, Infrastructure as a composed of physical and virtualized computing,
Service cloud, Optimization
storage and networking resources [5]. Taking storage as
I. INTRODUCTION an example, when a user uses the storage service of
cloud computing, he just pays for the consuming part
Cloud computing is defined as a new way of computing without buying any disks or even knowing anything
dynamically scalable and virtualized resources which about the location of the data he deals with. Sometimes
are provided as a service over the internet. It is a model the IaaS is also called Hardware-as-a-Service (HaaS)
for enabling on-demand network access to a shared pool [6].
of resources like servers, storage, which provides the
services that can be provisioned and released with Platform layer also called Platform-as-a-Service
minimal management effort [1]. Cloud computing generally abstracts the infrastructures and supports a set
represents a recent trend in IT that moves computing and of application program interface to cloud applications. It
data away from desktop into large data centres. It is an is the middle bridge between hardware and application.
application delivered as services over the Internet [2]. Examples of platform-as-a-services are Google App
The computing power in a cloud computing Engine and Microsoft’s Azure Services Platform [6].
environments is supplied by a collection of data centres, Application layer or Software-as-a-Service replaces the
in many different locations and interconnected by high applications that are running on the computer. If you are
speed networks [3]. In cloud computing, a cloud is a using SaaS then there is no need to install and run the
cluster of distributed computers which provides on- special software on your computer. Instead of buying the
demand computational resources or services to the software with higher cost, you just follow the pay-per-
remote users over a network [4]. use pattern which can reduce you total cost [5].
The resource management mechanism helps to
coordinate IT resources in response to management
actions performed by both cloud consumers and cloud
providers. It is the allocation of resources from resource
providers to resource consumers. Resource management
allows to dynamically re-allocatingresources, so that
user can more efficiently use available capacity.
In cloud computing, Resource Allocation (RA) is the
process of assigning available resources to the needed
cloud applications over the internet. IaaS cloud allocates Fig 1: Architecture for cloud computing

_______________________________________________________________________________________________
ISSN (Print): 2278-5140, Volume-3, Issue-2, 2014
31
 International Journal of Advanced Computer Engineering and Communication Technology (IJACECT)
_______________________________________________________________________________________________

B. Deployment Models hypervisor is useful both in private and public

clouds to gain performance improvements [7].
The Cloud model promotes four deployment models:
1) Private Cloud:The Cloud infrastructure is operated II. RESOURCE ALLOCATION
merely for an organization. It may be managed by In cloud computing, Resource Allocation (RA) is the
the organization or a third party and may exist on process of allocating available resources to the needed
premise or off premise. cloud applications over the internet. Resource allocation
2) Community Cloud: The Cloud infrastructure is is one of the demanding issues in cloud computing,
shared by several organizations and supports a where rare resources are distributed. From a consumer's
specific community that has shared concerns (e.g., viewpoint, resource allocation relates how services are
mission, security requirements, policy, and spread in the core of users. The benefit of resource
compliance considerations). It may be managed by allocation is that user neither has to install software nor
the organizations or a third party and may exist on hardware to access the applications, to develop the
premise or off premise. application and to host the application over the internet.
3) Public Cloud: The Cloud infrastructure is made
available to the general public or a large industry There is no limitation of place and medium. User can
group and is owned by an organization selling reach our applications and data anywhere in the world,
Cloud services. on any system.
4) Hybrid Cloud: The Cloud infrastructure is a Cloud providers can share their resources over the
composition of two or more Clouds (private, internet during resource scarcity.
community, or public) that remain unique entities
but are bound together by standardized technology Four different modes of hiring the computing capacities
that enables data and application portability [4]. from a cloud provider have been considered:

C. Virtualization 1) Advance Reservation (AR): Resources are reserved

in advance. They should be available at a specific
Virtualization technology provides the technical basis time
for Cloud Computing. In general, virtualization deals 2) Best-effort: Resources are provisioned as soon as
with the creation of virtual resources, such as operating possible. Requests are placed in a queue.
systems, servers, or storage devices. 3) Immediate: When a client submits a request, either
Different kinds of virtualization: System virtualization the resources are provisioned immediately, or the
adds a hardware abstraction layer on top of the request is rejected, based on the resource
hardware, which is called hypervisor or virtual machine availabilities.
monitor. Virtual Machines do not have direct access to 4) Deadline sensitive: assumed to be pre-emptible but
the hardware. The hypervisor runs virtual machines in a there is a limitation to their preemptability. It is
non-privileged environment. Using system pre-emptible only if the scheduling algorithm of
virtualization, multiple virtual machines, which may run Haizea can assure that it can be completed before
various operating systems, can be run on a single its deadline
physical machine. System virtualization is the very III. RELATED WORK
important technology that is used to provide IaaS, PaaS
and SaaS resources. A Virtual Machine Monitor JankiAkhani, [8]extendedtheadvance reservation
(VMM), also called as hypervisor, is asoftware that algorithms in Haizea to provide negotiation based
securely partitions the resources of a computer system allocation. Proposed algorithm generate counter offers
into one or more virtual machines. A guest operating provides counter offers considering constraints
system is an operating system that runs under the control flexibilities to maximize the chances of their acceptance.
of a VMM rather than directly on the hardware. The Using ranking algorithm, consumers will get suitable
VMM runs in kernel mode, whereas a guest OS runs in sorted offers according to their needs. These algorithms
user mode. Different hypervisors support different will not handle the situations when system has multiple
aspects of the cloud. requests for a single slot.
Hypervisors come in several types: Amit Nathani, [9]proposed an algorithm, which in
addition to pre-emption, also applies two concepts i.e.
1) Native hypervisors which sit directly on the
Swapping & Backfilling. It requires rescheduling of
hardware platform are most likely used to gain
fewer leases compared to the existing algorithm of
better performance for individual users.
Haizea. The proposed algorithm finds multiple slots in
2) Embedded hypervisors are integrated into a addition to finding single slot while scheduling a
processor on a separate chip. Using this type of deadline sensitive lease. An algorithm can developed to
hypervisor is how a service provider gains enhance the response time of best-effort leases.The
performance improvements. backfilling algorithm is just proposed it is not
implemented and tested and it can be considered as a
3) Hosted hypervisors run as a distinct software layer part of future work.
above both the hardware and the OS. This type of
_______________________________________________________________________________________________
ISSN (Print): 2278-5140, Volume-3, Issue-2, 2014
32
 International Journal of Advanced Computer Engineering and Communication Technology (IJACECT)
_______________________________________________________________________________________________

Fig 2: Resource Allocation Strategies in Cloud Computing

allocation policies in IaaS clouds. They first introduce

N.R. Ram Mohan, E.Baburaj, [10]estimates various
taxonomy of both types of policies, based on the type of
network resource allocation strategies and their
information used in the decision process, and map to this
applications in Cloud Computing Environment. A brief
taxonomy eight provisioning and four allocation
description for network resource allocation in Cloud
policies. Then they analyze the performance and cost of
Computing, based on differentially adapted dynamic
these policies with the developed framework SkyMark.
proportions, has also been done. In this paper authors
also addresses and categorizes the challenges normal to Hemant Kumar Mehta,Eshan Gupta [12]developedan
the resource allocation progress of cloud computing in economy based leasing algorithm and integrated with
terms of different types of resource allocation Haizea. This economy based algorithm takes care of
techniques. both the parties i.e. customer and service provider. It
uses the concept of optimization techniques to optimize
David Villegas [11]presented a comprehensive and
the costs. The incentive for customers is lower cost of
empirical performance cost analysis of provisioning and
_______________________________________________________________________________________________
ISSN (Print): 2278-5140, Volume-3, Issue-2, 2014
33
 International Journal of Advanced Computer Engineering and Communication Technology (IJACECT)
_______________________________________________________________________________________________

execution of its lease on the capable node as compare to providers, and want to maximise the profit earned
existing non-economy based algorithms. If an through serving SaaS users’ requests.
appropriate resource is not found, then the algorithm
Chunlin Li, Layuan Li[16] proposed the composition of
uses negotiation on budget and resource demand that
different layers in the cloud such as IaaS and SaaS and
increases the number of accepted lease. The proposed
its joint optimization for efficient resource allocation.
algorithm schedules lease based on its budget and cost
The efficient resource allocation optimization problem is
of resource along with the capabilities.
conducted by sub-problems. Through an iterative
Ahmed Shawish and Maria Salama[6]provided overview algorithm the proposed cloud resource allocation
on the Cloud’s anatomy, definition, characteristic, optimization algorithm is achieved.
affects, architecture, and core technology. They clearly
Jiayin Li, MeikangQiu [17], propose an adaptive
classify the Cloud’s deployment and service models,
resource allocation algorithm for the cloud system with
providing a full description of the Cloud services
pre-emptable tasks in which algorithms adjust the
vendors. They also addressed the customer-related
resource allocation adaptively based on the updated of
aspects such as the Service Level Agreement, service
the actual task executions. Adaptive list scheduling
cost, and security issues. Finally, it covers detailed
(ALS)and adaptive min-min scheduling (AMMS)
comparisons between the Cloud Computing paradigm
algorithms are used for task scheduling which includes
and other existing ones in addition to its significant
static task scheduling, for staticresource allocation, is
challenges.
generated offline. The online adaptiveprocedure is use
Gihun Jung [13]proposedan adaptive resource allocation for re-evaluating the remaining static resourceallocation
model that allocates the consumer’s job to an repeatedly with predefined frequency. In each re-
appropriate data center. The method to adaptively find a evaluationprocess, the schedulers are re-calculating the
proper data center is based on two evaluations: 1) the finish timeof their respective submitted tasks, not the
geographical distance (network delay) between a tasks that are assign tothat cloud.
consumer and data centers 2) the workload of each data
Zhenzhong Zhang [18] proposed resources allocation
center. The proposed model allocated the consumer
method named SLB to solve the problem of load
requests to the data center that is closed to consumer
imbalance in cloud environment. SLB consists of two
location.The proposed model can be improved to not
parts: (1) Online VM’s performance data statistical
only CPU-bond job with static workload, but also
analysis and resource demand forecast; (2) An algorithm
different size of memory space, dynamic workloads and
for the purpose of load balancing, which chooses proper
live migration situation. Furthermore, the proposed
host in resource pool based on the resource demand
model can be implemented in a real VMM environment
forecast of VM and the historical load information of
such as Xen Hypervisor.
hosts.The scope of statistical resources to memory usage
Yanbing Liu [14] proposed a new resource scheduling and disk I/O throughput, and further investigate the
model by introducing loyalty-based trust mechanism dynamic load balancing while VM running can be
into cloud computing. The paper proposes Cloud expanded
computing is a distributed computing system built on
top of cheap and unreliable hardware. The unreliability IV. CONCLUSION AND FUTURE WORK
of hardware requires that upper software. The model Cloud computing technology is progressively being used
gives assessment according to the real-time condition of in enterprises and business markets. The advancement of
the system and then allocates resources based on the cloud computing is intensely changing the perspective of
assessment. This kind of dynamic feedback mechanism information technology and eventually turns the utility
guarantees the stability of the system and reliability of computing into a reality. In cloud paradigm, an effective
the services effectively. resource allocation approach is required for achieving
Chunlin Li and Layuan Li, [15] provided an optimal user satisfaction and maximizing the profit for cloud
approach to satisfy cloud user requests and schedule service providers. Numerous techniques for resource
resources within cloud computing environments. CRA allocation have been proposed and implemented so far
algorithm is proposed by them. CRA algorithm works in the world of cloud computing. Resource allocation
on two levels i.e interaction between SaaS user and SaaS being one of the most critical factors which defines the
provider at the application layer and interaction between overall efficiency of the cloud has been solely based on
SaaS provider and cloud resource provider at the the technical parameters in the techniques discussed so
resource layer. This paper formulates SaaS solution as far. But cloud computing being a rapidly expanding
multiple optimal problems. The author considers business in the IT industry must also involve the use of
constraints such as SaaS user budget and deadline, SaaS parameters related to consumer context for resource
provider profit, and cloud resource provider profit, as allocation. This approach could result in a better overall
well as viewing the SaaS provider as both as resource performance and consumer satisfaction for the same.
provider and consumer. SaaS providers aim at
minimising the payment of using VMs from cloud

_______________________________________________________________________________________________
ISSN (Print): 2278-5140, Volume-3, Issue-2, 2014
34
 International Journal of Advanced Computer Engineering and Communication Technology (IJACECT)
_______________________________________________________________________________________________

REFERENCES (CICN), 2012 Fourth International Conference

on , 3-5 Nov. 2012
[1] Munich, Gerald kaefer “Cloud Computing
Architecture” IEEE Spectrum, February 2009 [11] Villegas, D.; Antoniou, A.; Sadjadi, S.M.; Iosup,
A., "An Analysis of Provisioning and Allocation
[2] Dikaiakos, M.D.; Katsaros, D.; Mehra, P.; Pallis, Policies for Infrastructure-as-a-Service
G.; Vakali, A., "Cloud Computing: Distributed Clouds," Cluster, Cloud and Grid Computing
Internet Computing for IT and Scientific (CCGrid), 12th IEEE/ACM International
Research," Internet Computing, IEEE , Sept.-Oct. Symposium on , 13-16 May 2012
2009
[12] Hemant Kumar Mehta and Eshan Gupta.
[3] Talib, A.M.; Atan, R.; Abdullah, R.; Azrifah, M., Economy Based Resource Allocation in IaaS
"CloudZone: Towards an integrity layer of cloud Cloud. Int. J. Cloud Appl. Comput. , April 2013
data storage based on multi agent system
architecture," Open Systems (ICOS) IEEE [13] Gihun Jung; Kwang-Mong Sim, "Agent-Based
Conference on , 25-28 Sept. 2011. Adaptive Resource Allocation on the Cloud
Computing Environment," Parallel Processing
[4] Dan C. Marinescu, “Cloud Computing Theory Workshops (ICPPW), 40th International
and Practice”, Elsevier, 2013 Conference on , 13-16 Sept. 2011
[5] Seyyed Mohsen Hashemi, Amid KhatibiBardsiri, [14] Yanbing Liu,Shasha Yang, Qingguo Lin and
“Cloud Computing Vs. Grid Computing”, ARPN Gyoung-Bae Kim “Loyalty-Based Resource
Journal of Systems and Software ,May 2012. Allocation Mechanism in Cloud Computing”,
[6] Ahmed Shawish and Maria Salama,”Cloud Recent Advances in CSIE, Springer, 2012
Computing: Paradigms and Technologies” [15] Chunlin Li and Layuan Li, “Resource allocation
Springer, 2014 in cloud computing: model and Algorithm”,Int. J.
[7] Miss. Rudra Koteswaramma, “Client-Side Load Web and Grid Services, 2013
Balancing and Resource Monitoring in Cloud”, [16] Chunlin Li • Layuan Li, “Efficient resource
International Journal of Engineering Research allocation for optimizing objectives of cloud
and Applications, November- December 2012 users, IaaS provider and SaaS provider in cloud
[8] JankiAkhani, Sanjay Chuadhary and Gaurav environment”,Springer, 2013
Somani, “Negotiation for Resource Allocation in [17] Jiayin Li, MeikangQiu, Jian-Wei Niu, Yu Chen,
IaaS Cloud”, Bangalore Compute Conf., pp 15 Zhong Ming, “Adaptive Resource Allocation for
ACM, 2011 Pre-empt able Jobs in Cloud Systems,” in 10th
[9] Amit Nathani, Sanjay Chaudhary, and Gaurav International Conference on Intelligent System
Somani, Policy based resource allocation in IaaS Design and Application, Jan. 2011.
cloud. Future Gener. Comput. Syst. , 2012 [18] Zhenzhong Zhang; Haiyan Wang; Limin Xiao; Li
[10] Mohan, N.R.R.; Raj, E.B., "Resource Allocation Ruan, "A statistical based resource allocation
Techniques in Cloud Computing -- Research scheme in cloud," Cloud and Service Computing
Challenges for Applications, " Computational (CSC), International Conference on ,12-14 Dec.
Intelligence and Communication Networks 20


_______________________________________________________________________________________________
ISSN (Print): 2278-5140, Volume-3, Issue-2, 2014
35