A Light-Weight Operating System For IoT Devices

A Light-Weight Operating System for
Internet of Things Devices
Emekcan Aras
Embedded Computing Systems

Submission date: June 2016
Supervisor: Amund Skavhaug, ITK
Norwegian University of Science and Technology

Department of Engineering Cybernetics
A Light-weight Operating System for Internet
of Things Devices
Emekcan Aras
June 2015
PROJECT / MASTER THESIS
Department of of Engineering Cybernetics
Norwegian University of Science and Technology
Supervisor: Associate Professor Amund Skavhaug

i
Preface
This master thesis was written the spring of 2016 at the Norwegian University of Technology
and Science, Departmen of Engineering Cybernetics. It has been a part of my master program
EMECS( European Master in Embedded Computing Systems) and this thesis concludes my Mas-
ter’s degree.
This Master’s thesis is a contribution and new way of look to embedded operating systems.
It is not the continuation of the specialization project of the fall semester 2015 even if the overall
goal of developing the robust wireles networks still stands. This thesis has not been submitted
previously and has been made by independent work. In addition, a conference paper is being
written based on this paper which focus more on IOT service layer.
The thesis can be divided into two important parts. The first part is developing a very small
“microscopic” operating system. The design steps and background lies behind the design is
explained in detail in this particular paper.
The second part is radio communication protocol and IOT service layer which is presented
as the novelty of the project. Since there are many different operating system and architec-
ture can be found in the market, we have revealed that a specific embedded operating system
designed for IOT and wireless sensor networks that enables resource and information sharing,
would be a good research topic. Therefore, the particular project can be considered as the col-
lection of the related background education, embedded software skills, hours of discussions and
sparks of new ideas about embedded operating systems.
This particular paper has been written for three types of readers. The one who would like
to read and learn something about operating systems and embedded system, the students who
would like to continue the project and the professors who will asses this paper.
Trondheim, June 2016
Emekcan Aras
ii
Acknowledgment
I have now spent two years in abroad since I left Istanbul for this Joint Degree program. I have
lived in two beautiful cities Southampton and Trondheim and I have found to chance to meet
with wonderful people in these years. I will soon be graduated and leave Trondheim but I will
never forget this period of my life.
I would like to thank my friends in EMECS program who have worked hard with me and
never hesitate to help me for anything. In addition, I would like to thank my beloved friends
in Istanbul, Glasgow, Berlin and Milano who gave me the inspiration to apply this program and
supported me during it, we may live away from each other but they will never be forgotten.
I would like to thank my supervisor professor Amund Skavhaug for the help, guidance and
invaluable ideas he has given during the semester.
Lastly, I would like to thank my beloved father and mother who always supported me during
this program and taught me that another world without any war or discrimination is possible.
E.A.
iii
Summary and Conclusions
Nowadays, all around of us has been surrounded by wireless devices. More than a decade, engi-
neers have been using these wireless devices in industry to measure environmental data. Large
networks, which are formed by hundreds or thousands sensor nodes, are not easy to build,
maintain and observe. Moreover, with the trending concept called Internet of Things, devices
have gained different functionalities and the different type of networks started to develop. In
order to design an efficient and low cost sensor and IOT network, a different type of architecture
in both network side and low-level software side must be developed.
Designing a specific operating system motivated us earlier part in this project. After detailed
investigation, it is revealed that resource-constrained embedded devices need a new type of
embedded operating system specifically developed for sensor and IOT network. Moreover, in
the market, there are many choice for embedded operating systems. However, an open-source
operating system for embedded devices in the sensor and IOT networks is needed and could be
useful for educational purposes as well as commercial purposes.
In the light of the motivations that is mentioned at the paragraph above, a light-weight em-
bedded operating system has been designed and developed from scratch in this thesis. In addi-
tion to that, a radio protocol and an IOT service layer which act as middleware in the network
have been designed and implemented into the embedded operating system. All the background
lies behind this project and the designing steps are explained in detail in this thesis. In addition
to this, a conference paper about the operating system particularly IOT service layer is being
written.
Keil uVision5 software development platform was used to develop the mentioned operating
system in this project. Detailed investigation and research were done about embedded systems,
operating systems, distributed systems and wireless communication before the project. After-
wards, entire software libraries for the kernel and hardware drivers were developed from scratch
by using C and Assembly language. Since the project is focused on embedded devices, one of
the low-power processor architectures called ARM Cortex-M0 architecture has been used as the
target platform. An SOC from Nordic Semiconductor called NRF51822, which contains this pro-
cessor architecture, is used inside of the devices in this project. In order to test and validate the
iv
results and functionality of the system, two different development kits have been used provided
by the vendor company. Detailed tests scenarios have been tried and a small working sensor
network/cluster has been formed by using these devices and the operating system.
At the end of the project, a small size operating system which requires around 8Kb of flash
memory and 7Kb of RAM, has been created. Tasks which is given to devices has been efficiently
executed without causing any overhead or failure. In addition to this, functionality which is
provided by the IOT layer has been achieved as well. Two types of device are introduced with
the proposed IOT layer. A standard component for sensor and IOT networks called “node de-
vice” and supervisor component of the network called “manager device” are tested. After tests
and verifications, it has been seen that resource and information sharing across the network is
achieved.
In this thesis project, resource-constrained low profile embedded devices and architectures
selected as the target device group. It is revealed that using a specifically design operating sys-
tem, these types of devices could be used in many applications with reasonable prices. More-
over, the proposed operating system layer could bring a new way of thinking on the sensor and
IOT device networks and could be used as the solution for problems and challenges. Further-
more, with the user-friendly design particular operating system could be very useful in educa-
tional purposes.
List of Abbreviations and Symbols
IOT Internet of Things
RAM Random Access Memory
CORBA Common Object Request Broker Architecture
SOA Service Oriented Architecture
OSI Open System Interconnection
OS Operating System
API Application Porgramming Interface
HAL Hardware Abstraction Layer
CPU Central Processing Unit
TCP IP Transmission Control Protocol-Internet Protocol
SOC System on Chip
TI Texas Instruments
RTOS Real-Time Operating System
ROM Read-Only Memory
GUI Graphical User Interface
IPC Inter-process Communication
v
vi
RF Radio Frequency
VLF Very Low Frequency
EHF Extremely High Frequency
ASK Amplitude Shift Keying
PSK Phase Shift Keying
FSK Frequency Shift Keying
PRN Pseudo-Random-Noise
DSSS Direct Sequence Spread Spectrum
FHSS Frequency Hopping Spread Spectrum
ISM Industrial Scientific Medical
IO Input-Output
GPIO General Purpose Input-Output
PCB Printed Circuit Board
RSSI Received Signal Strength Indication
AES Advanced Encryption Standard
CRC Cyclic Redundancy Check
DMA Direct Memory Access
I2C Inter-Integrated Circuit
SPI Serial Peripheral Interface
PSP Process Stack Pointer
MSP Main Stack Pointer

vii
ISR Interrupt Service Routine
SVC Service Call
RISC Reduced Instruction Set Computing
MIPS Microprocessor without Interlocked Pipeline Stages

List of Figures
1.1 Prediction about IOT devices [5]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 A survey about Internet of Things 1 [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 A survey about Internet of Things 1 [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 A general structure of Embedded Systems [9]. . . . . . . . . . . . . . . . . . . . . . . 12

2.2 Monolithic System Architecture [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Layered System Architecture [13]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Microkernel Architecture Structure [12] . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Difference between Client-Server and Multi-Tier Architecture . . . . . . . . . . . . 20
2.6 Implementation of CORBA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.7 Structure of SOA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.8 OSI model [19] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.9 Frequency Ranges [20] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.10 Types of Digital Modulation [21] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.11 Spread Spectrum Communication [22] . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.12 Overview of Wireless Technologies in the market [24] . . . . . . . . . . . . . . . . . . 29
3.1 Kernel Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.2 Structure of OS Services with API and Application Layer . . . . . . . . . . . . . . . . 39
3.3 Proposed Operating System Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Structure of IOT Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Block Diagram of NRF51822 [34] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Nrf51822 Radio Module Block Diagram [35] . . . . . . . . . . . . . . . . . . . . . . . 47
viii
LIST OF FIGURES ix
4.1 Structure of Hardware Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2 Interaction between HAL and Hard Drivers and Structural Design . . . . . . . . . . 54
5.1 Multitasking and Context Switching[37] . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.2 Context Switching from one task to another[37] . . . . . . . . . . . . . . . . . . . . . 57
5.3 Context Switch without PendSV[37] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.4 Flow diagram of system tick handler and dispatcher . . . . . . . . . . . . . . . . . . 61
5.5 Transition of Task States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.6 The flow chart of priority-driven scheduler during the context switching . . . . . . 65
6.1 Structure of Linked List [38] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

6.2 Main memory after the memory allocation [39] . . . . . . . . . . . . . . . . . . . . . 70
6.3 The structures and functionality of API . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1 Structure of Soft-Device[41] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

7.2 On-air packet layout [35] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.3 Structure of Radio Library and Interactions . . . . . . . . . . . . . . . . . . . . . . . . 80
7.4 Packet Layout of the New Radio Protocol . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.5 The stack structure of the radio protocol based on OSI model . . . . . . . . . . . . . 82
8.1 Classical Cluster Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

8.2 Cluster with Main Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8.3 Cluster with Nodes Net . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.4 IOT service software structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.5 Structure of Mailing System and Yellow Pages . . . . . . . . . . . . . . . . . . . . . . 93
8.6 Flow Chart of Registration Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
9.1 NRF518222 Development Kit and NRF51822 Dongle . . . . . . . . . . . . . . . . . . 100

9.2 Structure of Test Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
List of Tables
2.1 Evaluation of Embedded Operating Systems . . . . . . . . . . . . . . . . . . . . . . . 33

2.2 Comparison of Operating Systems Embedded . . . . . . . . . . . . . . . . . . . . . . 34
8.1 Pros and Cons of Suggested Strcutres. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

8.2 Possible Problems and Solutions of Suggested Structures . . . . . . . . . . . . . . . 90
9.1 Size of the Embedded Operating System . . . . . . . . . . . . . . . . . . . . . . . . . 102

9.2 Execution Time of IOT Service Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
x
Listings
4.1 Hal structure for GPIO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.2 HAL structure of communication drivers . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1 Task Control Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.2 Mutex Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.1 Configuration File of Particular Operating System . . . . . . . . . . . . . . . . . . . . 74
8.1 IOT node resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.2 IOT node resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
xi
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
1 Introduction 2
1.1 Internet of Things and Embedded Systems . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Example Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Challenges and Problems in IOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 A specific Operating System for Internet of Thing Devices . . . . . . . . . . . . . . . 8
2 Background Research 11
2.1 Operating Systems and Real-Time Systems . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Distributed Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Wireless Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3 Details and Features 35

3.1 Simple and Light-weight Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Two Dimensional Kernel Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Internet of Things Service and Resource Sharing . . . . . . . . . . . . . . . . . . . . . 41
3.4 3.4. Development Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4 Hardware Abstraction Layer and Drivers 48
5 Kernel Architecture 55
xii
CONTENTS 1
5.1 5.1. Multitasking and Context Switching . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.2 Priority-Driven Scheduler and Real-Time Compatibility . . . . . . . . . . . . . . . . 62
6 Other System Components 67

6.1 Utilities Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.2 Application Programming Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.3 User defined functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7 Radio Library, Stack and Protocol 76
8 IOT Service and Security 84

8.1 Network Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.2 IOT Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.3 Security Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
9 Tests and Results 99
10 Discussion, Future Work, and Conclusion 103

10.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
10.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
10.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
11 Bibliography 116
Chapter 1
Introduction
1.1 Internet of Things and Embedded Systems
In last twenty years, many technologic achievements have been happened in the Integrated
Circuit industry. These have led to develop small, energy efficient and faster processors and
microcontrollers. Nowadays even small microcontrollers which cost ten to twenty dollars, have
enough CPU power and memory space to achieve complicated tasks such as managing with
TCP/IP stack or Bluetooth communication. Even some operating systems can be supported by
these kind of low cost microcontrollers. Because of these, every kind of embedded systems have
started to connect with each other and internet. This concept is called Internet of things (IOT).
Internet of Things represents the devices, which communicates and shares data among each
other using different protocols [1]. In 1993, a system developed in University of Cambridge to
help people working in other parts of the building, avoid pointless trips to the coffee pot by
providing, on the user’s desktop computer, a live 128×128 greyscale picture of the state of the
coffee pot. Since it was real time and online application [2], this system is considered as first
example of internet of things. However the name of the concept actually used by Kevin Ashton
in 1999 in a presentation, which was given by him to a private cooperation [3]. However, this
concept has gained broader meaning during the years. Nowadays Internet of Things is defined
as “a pervasive and ubiquitous network which enables monitoring and control of the physical
environment by collecting, processing, and analyzing the data generated by sensors or smart
objects.” [4]
2
CHAPTER 1. INTRODUCTION 3
Figure 1.1: Prediction about IOT devices [5].
As it can be seen from the figure 1.1, for almost 15 years, we as human kind have been trying
to connect everything to each other and to Internet. The estimations say that in 2020 there will
have been 50 billion IOT object in the World. That means there will be six device per person.
Therefore, it has already become dominant concept for the industry. In order to implement this
concept, new approaches in communication and software had been started to investigate and
develop.
The semiconductor vendors have already started to produce architectures desired by the
market. System on Chip devices are becoming more popular day by day. A system on a chip
or system on chip (SoC or SOC) is an integrated circuit (IC) that integrates all components of a
computer or other electronic system into a single chip. It may contain digital, analog, mixed-
signal, and often radio frequency functions all on a single chip substrate. The contrast with a
microcontroller is one of degree. Microcontrollers typically have under 100 kB of RAM (often
just a few kilobytes) and often really are single-chip-systems, whereas the term SoC is typically
used for more powerful processors, capable of running software such as the desktop versions of
Windows and Linux, which need external memory chips (flash, RAM) to be useful, and which
are used with various external peripherals. In short, for larger systems, the term system on a chip
is hyperbole, indicating technical direction more than reality: a high degree of chip integration,
leading toward reduced manufacturing costs, and the production of smaller systems. Many
systems are too complex to fit on just one chip built with a processor optimized for just one of
the system’s tasks. SOCs are very common in the mobile electronics market because of their
low power consumption. [6] These developments in both silicon industry and wireless network
industry have led to new communication protocols and new low cost small processor in the
market. System on Chip devices with the radio chip has been introduced to the market. In
addition, the protocol stacks for ZigBee, Bluetooth Low Energy, ANT, and Thread have been
implemented inside of those SOCs.
1.2 Example Applications
Together with embedded (resource constrained) devices, many different applications have been
developed and implemented on many different fields.
• Automotive Industry: Consumers want the digital experiences in their vehicles to align
with the ones they enjoy everywhere else. When tied to the IoT, the car is an integral part
of the interdependent web of information flow, turning data into actionable insight both
inside the car and in the world around it
• Energy Industry: Through the IoT, the power grid’s countless devices can share informa-
tion in real time to distribute energy more efficiently. Consumers, businesses, and utility
providers get the information they need to better manage their energy-connected things
to consume less energy.
• Fitness and Environment: Fitness bands, watches, and even smart clothes are able to
monitor and transmit data on your daily activity levels through step counting, heart rate
and temperature.
• Health: These wearables monitor crucial health factors like oxygen saturation, heart rate
and more, and can communicate any results outside of a programmed range to the patient
and to her physician.
• Retail: Retailers use the IoT to provide personalized and immersive experiences that keep
shoppers coming back. Gathering and organizing data is only part of the challenge.
• Smart Manufacturing: The benefits of IoT products include software and hardware side
that ease and accelerate design time for smart manufacturing application.
• Smart Buildings: The IoT is enabling a transformation in building efficiency and manage-
ment.
• Home Automation: From enhancing security to reducing energy and maintenance costs,
there are a many of innovative IOT technologies for monitor and control of smart homes.
1.3 Challenges and Problems in IOT
As it was mentioned earlier in this paper, novel architectures in digital design and networks pro-
tocols make the development of Internet of Things networks easier. All around us surrounded
by smart objects. Moreover, most of the people have started to work in smart buildings, live
in smart cities. In our daily life, we use devices to measure and track our health related most
private data such as heart rate, blood pressure and even our location.
On the other hand, the industry has realized the importance of IOT objects as well. A concept
called Industry 4.0 is becoming more important day by day for the industry. Industry 4.0 is a col-
lective term embracing a number of contemporary automation, data exchange and manufactur-
ing technologies. It had been defined as a collective term for technologies and concepts of value
chain organization, which draws together Cyber-Physical Systems, the Internet of Things and
the Internet of Services. [7] In industry, large sensor networks and smart objects have started to
use in their production line. Demand of IOT devices in industry has been increasing for last ten
years. At the figure 1.2 , needs of industry can be seen.
Secure devices is the most important need together with easy integration. In addition, shar-
ing device resource/data and remote monitoring/management of the devices can be considered
as second most important features as well. In order to develop embedded IOT devices according
to needs of industry, we as developers need to identify the problems and challenges and find to
way to overcome.
The figure 1.3 gives the perspective from developer and producer point of view. According
to developers integration of the devices and total cost are the most difficult hurdles to overcome.
Figure 1.2: A survey about Internet of Things 1 [8].
Moreover, to develop hundred percent secure devices is also one of the noticeable challenges.
Furthermore, since low cost embedded devices have turned out to be the best solution for inter-
net of things as well as the sensor networks, large networks, which may be formed by hundreds
or thousands of nodes, can be created easily with these devices. However, in the application
side, this new trend brought some new problem to the networks and software. Since wireless
communication not reliable as wired communication, there are always some minor errors such
as missing packet etc. in the wireless network. In addition to that, most of the sensor network
are used in critical measurements in the harsh environments. Sensor nodes may become un-
available because of lots of different reasons. Sometimes these failure may not create big prob-
lems. In addition, we started to use devices, which can collect our private data such as heart
rate, blood pressure, location etc. We make everything accessible. That also caused problems
in security side. Together with other problems, internet of things devices and networks could
be unsecure and inefficient. In order to prevent devices and network such scenarios, a new ap-
proach just for the internet of things devices must be developed in both the operating system ,
software and network side. Challenges must be detailed and identified better to develop a novel
embedded Internet of Things devices. Challenges and difficulties can be divided into three part.
These parts are embedded software/operating system, hardware/resource constrained and net-
work related challenges.
• Software/Operating System Challenges;

Figure 1.3: A survey about Internet of Things 1 [8].
– Integration and implementation
– Development of Real Time Kernel
– Resource sharing
– Secure and upgradable Operating System
• Hardware Challenges;
– Memory Constrain( hard to build an Operating System inside of small microcon-

troller)
– Limited Encryption Capabilities
– Energy Consumption and total cost
• Network Challenges;
– Implementation of Large Networks( hard to set-up thousands of nodes)
– Lack of global standard
– Fault detection and avoiding network failure design for Internet of Things
In this particular project, a specific Operating System design for Internet of Things devices is
proposed as a solution to overcome many of these challenges and problems. Specifications and
an overview of the proposed operating system are mentioned in the following chapter.
1.4 A specific Operating System for Internet of Thing Devices
After identifying problems and challenges about embedded devices on Internet of Things net-
works, a specific operating system for IOT devices had been decided to be designed. Since low
power and low cost microcontroller has become most suitable architecture for embedded IOT
devices, proposed architecture specifically will be developed in this kind of platform. More in-
formation about platform and microcontroller, which has been used during this project, is de-
tailed later on this paper.
Since our focus is on memory constrained embedded devices, firstly the kernel has to be
minimized to have a successful implementation. In order to achieve minimization, the dis-
patcher and the scheduler should be minimized. In addition, the operating system services
must be reduced to save memory space. Most of the operating system have large footprints be-
cause of the number of services implemented on operating system. However specific applica-
tions for sensor networks and Internet of Things do not require complicated operating systems.
Secondly, it should have priority driven scheduler to make them real time compatible. Real-time
tasks could have higher priority to not miss the deadlines. In addition, low power consumption
should be taken in consideration. Different power modes must be introduced. Moreover, idle
function must be designed carefully to achieve this purpose.
However doing optimization in classical design in embedded operating systems is not enough
to overcome all challenges and problems earlier specified in this paper. A new perspective is
needed to achieve to develop such an operating system. A new kernel architecture specific for
IOT devices could help us to achieve easy integration and the user-friendly operating system.
In this particular project a new two dimensional kernel architecture is proposed. This archi-
tecture is also planned to be a light weight and implemented with priority-driven scheduler. A
new operating system service called IOT service is proposed. The aim of the service is to allow
resource sharing and easy integration with other devices without using any other library. This
service is planned to be out on top of every layer even application layer in the operating system
architecture. Therefore, users would need only develop an application for specific node. Since
IOT service would handle all integration and sharing, application development would be like
developing a basic firmware. Details about design and development of particular embedded
operating system is mentioned another chapter in this thesis.

In addition to a new kernel architecture and IOT service, device security must be achieve in
operating system level. Security should not be thought of as an add-on to a device, but rather
as integral to the devices reliable functioning. Software security controls need to be introduced
at the operationg system level, take advantage of hardware security capabilities. Most of the
embedded device use old operating systems such as TinyOS, FreeRTOS etc. However, none of
them was developed specific for IOT. A detailed comparison of embedded operating systems in
investigated later on this paper. Feature of proposed operating system can be seen at list below.
• Microscopic(light-weight) Kernel Architecture
– Reduced OS services
– Small footprint and Efficient Task execution
• Two dimensional Operating System
– User friendly
– Gives two different perspective and different system layers
• A specific Operating System service for Internet of Things application
– Enables resource sharing and distribute the workload of the network
– Easy integration
– Specifically developed for IOT devices
• Real-Time compatible
– Priority-driven scheduler
• Low energy consumption
– Different power modes for different applications
• Security in Operating System Level
– A new system layer for security

– Separated Communication functionally from application layer
The main contributions of this thesis are:
• A new kernel architecture for embedded operating system is proposed.
• Detailed Background Research about Operating Systems, Distributed Systems and Wire-
less Communication is done.
• Hardware Libraries, Kernel Functions and OS services are developed from scratch.
• A new radio protocol is and implemented.
• A new OS service called IOT service which acts as middleware and enables resource and
information sharing across the network, is introduced and developed.
• Using this embedded operating system, a cluster is formed and all the system functional-
ities are tested in real world.
The remaining Chapters of this thesis are organized as follows. Chapter 2 provides a lit-
erature and technologies review related with the operating and distributed systems, wireless
communication. In addition to that, it contains related work about the project topic. Chapter
3 details and features about proposed operating system and kernel architecture. Chapter 4 de-
scribes the implementation of the hardware abstraction layer. Additionally, it gives information
about hardware drivers, which are used by HAL. Chapter 5 focuses on the kernel architecture
and functionality. In addition to that, in chapter 6 information about other small system com-
ponents are given. In chapter 7 design and structure of a small radio library, stack and protocol
which is developed and implemented specifically for this operating system is explained. Our
novel operating system service called IOT service and network structure lies behind it are ex-
plained in detail in Chapter 8. The results are shown at Chapter 9 and lastly, the ideas discussed
along this thesis work are concluded in Chapter 10. Additionally, it provides the future work and
lessons learned during the development of this work. References and materials used during this
project can be found at chapter 11.
Chapter 2
Background Research
To develop such an embedded operating system, which is mentioned in earlier section, many
disciplines in computer science and electronics, must be investigated and researched. In this
section, related background research and topics are detailed. Firstly brief explanation about
embedded systems is given. Afterwards operating systems and real time system concepts are
explained in detail. In addition, since resource and information sharing is one of the features
of the particular operating systems distributed system topic is investigated. Moreover, to create
reliable, secure and fast communication between devices, fundamentals of wireless communi-
cation is mentioned from the embedded software and electronics point of view. At the last part
of the chapter, research has been done about related work and comparison of the embedded
operating system which are currently in the market, can be found.
“Last few decades have seen the rise of computers to a position of prevalence in human
affairs. It has made its mark in every field ranging personal home affairs, business, process au-
tomation in industries, communications, entertainment, defence etc...” [9]. Embedded systems
can be defined as a computer system which is designed and optimized to execute a dedicated
task and where the computer itself is not necessarily the purpose of the system. Any device that
contains a programmable computer which introduce some degree of intelligence to the system,
but which itself is not a general purpose computer (such as PCs, workstations, laptops, tablets).
An embedded system is generally a system within a larger system. Modern cars and trucks con-
tain many embedded systems. One embedded system controls anti-lock brakes, another mon-
itors and controls vehicle’s emission and a third displays information on the dashboard. Even
11
CHAPTER 2. BACKGROUND RESEARCH 12
Figure 2.1: A general structure of Embedded Systems [9].
the general-purpose personal computer itself is made up of numerous embedded systems. Key-
board, mouse, video card, modem, hard drive, floppy drive and sound card are each an embed-
ded system. For almost twenty years we as human being are surrounded by specific embedded
systems like smartwatches, cars, temperature sensors. Right now, even coffee machines, wash-
ing machines and other assets in our daily life, become smart embedded system with the help
of Internet of Things concept.
2.1 Operating Systems and Real-Time Systems
Every embedded platforms contain a processor and software to run the processor. Mostly there
must be a place to store the executable code and temporary storage for run-time data to run
the software. These take the form of ROM and RAM respectively. If memory requirement is
small, it may be contained in the same chip as the processor. However, there are also external
memory chips which one or both types of memory can be placed. In addition to memory also
every embedded system must have some kind of input and outputs to collect data from real
world and manipulate it accordingly [9]. For instance, in a washing machine the inputs are
the buttons on the front panel and temperature sensor or probe in water tank and the outputs
are the human readable display and rpm. Generally, inputs to the embedded system are in the
form of sensors and probes, communication signals, or control knobs and buttons. Outputs are
generally displays, different kind of signals and changes to the real world. General structure of
an embedded system can be seen at the figure 2.1.
1. Processor Management: An operating system must ensure that every single process and
task receives same amount of execution time from processor while using maximum pro-
cessor cycles to not miss hard deadlines. In addition, it is also responsible for switching
processes/task in multi-tasking environment [9].
2. Memory and Storage Management: Every task needs amount of memory to perform spe-
cific tasks properly. An operating system also allocates enough memory for task/processes
and uses different types of memory efficiently.
3. Device Management: All the hardware on the system is managed by the operating system.
The hardware drivers form a way for the user applications to use of hardware functionality
without the need for details about hardware operation [9]. “The driver’s function is to be
the translator between the electrical signals of the hardware subsystems and the high-level
programming languages of the operating system and application programs “[9]. Drivers
are abstracted from the operating system kernel for upgradability of devices.
4. Providing Common Application Interface: In order to achieve easy upgradable operating

system and security, abstraction layers is used in operating system. Hard drivers is the first
abstraction layer in the operating system and application program interface is second and
most important one. With the help of APIs, programmers for particular operating system
can use the functions/services of operating system without considering CPUs other oper-
ation. Usually operating system deals with the details about particular service/functions
when programmer/developer uses an API functions [10].
5. Providing Common User Interface: Usually in order to achieve interaction between user
and system there is a user interface. Although most of the large operating system like Mac-
intosh or Microsoft have graphical user interface, in embedded operating system mostly
no need to have such a GUI since embedded devices usually perform specific tasks and
no need to have user interaction.
In addition, an operating system is a compound structure of software. Therefore architecture of

the operating system can be built in many different ways. Some architecture about operating
systems design have been currently used in popular embedded operating system in market, are
mentioned the lines below.
Figure 2.2: Monolithic System Architecture [12].
1. Monolithic systems are the first and one of the oldest architecture. However, it is still
popular in the market especially very small real time applications for embedded systems.
Because it is very simple and easy to implement and it has very low processor and memory
overhead. [11]
In this architecture, all of the system runs in privileged mode. “The only internal structure
is usually induced by the way operating system services are invoked” [11]. In application
layer, which is run in user mode, can request operating system services using a special
instruction, mostly it is called system call instruction [11]. This “superior” call provide
privileged mode and transfers control to system call dispatcher of the operating system.
System call dispatcher controls and decides which service must be carried out and trans-
fers control to suitable procedure.
Handling of interrupts are done directly in the kernel (at least most of the part) and han-
dlers are not fully authorized processes or functions. Therefore, overhead which can be
caused by interrupt handling is very small since there is no full task switching at interrupt
arrival. However, interrupt handler (according to its priority) cannot run most of the op-
erating system services. In addition, scheduler of operating system is not running while
interrupt handler is running or in progress and also priority according the hardware inter-
Figure 2.3: Layered System Architecture [13].
rupt requests is in effect, therefore handler functions or task is run at a priority higher than
the priority of all other task and it may have higher priority than some system services. At
the figure 2.2, you can see the structure of Monolithic Systems.
2. Layered systems. “A refinement and generalization of the monolithic system design con-
sists of organizing the operating system as a hierarchy of layers at system design time.” [11]
Every single layer is built on top of the operating system services given by the one layer be-
low, hence mostly richer and well-described set of services to the layer above it. Memory
and processor overheads are similar to monolithic systems, since interrupt handling and
system interface implemented similarly to it.
Since structural design and modularity is the key part of this architecture, maintenance is
easier. Operating system code is easy to read and understand. Moreover, since structural
design can be changed without changing other layers, interlayer interfaces does not need
to be changed. Structural form of layered system can be seen at the figure 2.3 .
3. Microkernel systems. This design is based on to have smallest kernel as possible in the
operating system. Therefore most of the operating system services, they run in user mode
and only a few amount of core service can be run in privileged mode. There is a messaging
system between kernel and other services and tasks. All the processes and services in user
mode send a request message for core services to suitable operating system server and
wait for a reply. [11]
The key concept drives microkernel architecture to handle all the communication be-
tween applications and server is security. This kind of architecture enforce users to have
Figure 2.4: Microkernel Architecture Structure [12]
proper security policy on such communication since it would be difficult or inefficient to

perform some critical operating system functions in user-mode processes.
Moreover, also it help to make operating system easier to manage and maintain. In addi-
tion, using that kind of message passing system between processes and operating system
components makes system modular and it leads to have clear understanding on operating
system components.
There are some ways to design embedded and real-time software applications. One of the
convenient and easy to understand way to design it is organizing them like sequential set of
processes which cooperates with each other [11]. Processes are the key concept in every oper-
ating system.” A process is basically a program in execution. Associated with each process is
its address space, a list of memory locations from 0 to some maximum, which the process can
read and write” [10]. In addition to that, every process usually associated with set of resources,
including registers and all information needed to execute a program.
Moreover, usually operating systems supports to control multiple threads within the same
process, sharing the same address space. In the implementation of threads, user mode can used
for most of the part without the kernel intervention. Also the address space does not change its
contents and must not be switched to achieve very fast switching between threads with respect
to switching between processes. Besides, as it was mentioned before all the threads within a
process share the same address space, therefore there is limited protection among threads. Thus
for instance, a thread can block or disable or may cause error by mistake another thread registers
or data and for operating system, it is impossible to detect this kind of errors [14].
Thereupon, most of the embedded and small operating systems only implement threads
which can keep overheads and hardware requirements to a minimum, while operating system
for complex applications supports single and multiple process model to enhance the reliability.
Another important part of embedded and real-time operating systems is the scheduler. Task
of scheduler is to decide which runnable threads or task available in processor and for how
long. Schedulers can schedule threads and task beforehand that is called static schedulers and
they can perform scheduling during runtime which is called dynamic schedulers. There are
several different algorithms for dynamic schedulers however, since a priority driven scheduler
is implemented in particular project, dynamic scheduler algorithms are not mentioned in this
paper.
However sometimes different task and processes may need to work synchronize to each
other or use to same resources. To achieve this task an essential function of multi-programmed
operating system known as Inter-Process Communication must be implemented. Many inter-
process synchronization and communication mechanisms have been proposed and were ob-
jects of extensive theoretical study in the scientific literature. Since semaphore concept has
been used in this project, only this concept is detailed.
"A semaphore, first introduced by Dijkstra in 1965, is a synchronization device with an inte-
ger value, and on which the following two primitive, atomic operations are defined"[11]:
• The P operation, often called DOWN or WAIT, checks if the current value of the semaphore
is greater than zero. If so, it decrements the value and returns to the caller; otherwise, the
invoking process goes into the blocked state until another process performs a V on the
same semaphore.
• The V operation, also called UP, POST, or SIGNAL, checks whether there is any process
currently blocked on the semaphore. In this case, it wakes exactly one of them up, allowing
it to complete its P; otherwise, it increments the value of the semaphore. The V operation
never blocks.
The brilliance of semaphores is simplicity in implementation. In addition, this concept is a

very low level IPC mechanism. Therefore it helps to have a low processor and memory overhead
especially on uniprocessor systems. On the other hand, its low level feature causes difficult
usage in complex application. Especially they can easily cause to mutual exclusion which leads
the problem of priority inversion. If a higher priority process or task is stopped or waited by
lower priority process or task for shared resources, this leads undesired behaviour in scheduler
and called priority inversion. In order to avoid from this problem, most of embedded and real-
time operating system uses different kind of algorithms like priority inheritance, priority-ceiling
protocols, etc.
2.2 Distributed Systems
As it was mentioned before, purposed operating system enables resource and information shar-
ing between embedded devices. In order to achieve this task, distributed systems concept must
be investigated and implemented. A distributed systems is a software and hardware system in
which components located on some kind of network and communicate and coordinate their
actions by passing messages. There are three important characteristics of a distributed system :
1. Concurrency: Tasks or workload on the network can be done at the devices separately.
They can shared devices over network when they needed. Therefore, system power or
scale can be increased easily by just adding more devices to distributed network [15].
2. No global clock: Since all the communication done among devices is made by message,
there is no reason for a global clock to synchronize devices. Most of the other networks or
environments a global clock needed to synchronize devices [15].
3. Independent failures: Every device in the network (distributed systems) may fail and this
does not affect other components of the system. Therefore, distributed system can keep
running independent from device failure [15].
These characteristics are fundamental to the understanding of distributed systems. These

make a distributed system what it is. In addition, there are several reason to use this concept in
this particular project. These main reasons can be classified into six section.
1. Resource sharing: In distributed systems, resources can be shared easily. Not only soft-
ware resources but also hardware resources or data can be shared among the devices. This
is one the main factor to implement distributed system concept in this project.
2. Openness: Making system or network cluster components public to other networks is also
important purpose of distributed systems. This leads systems to be more extensible [16].
3. Scalability: An Internet of Things network based system must be scalable. With the help of
the concepts, adding more resources to the network becomes less costly and simple [16].
4. Fault-tolerance: Every task, device or network may fail independently of the other system
component. Shared resource can be installed on many devices in a distributed system
thus, handling with the loss of that particular resource if one device fail, is important.
Network failures or loss of a resource in the network can be detected using different algo-
rithms in middleware to make other components of the network more reliable [17].
5. Concurrency: Since there is no global clock and dependencies, each devices connected
to system can run concurrently. System can execute different tasks in different devices by
using messaging system [16].
6. Transparency: Last reason to implement distributed system concept is make system more
transparent to the user. This can be divided into three sub section.
(a) Location transparency: Local and remote resources can be accessed seamlessly.
(b) Failure transparency: Masking of failures can be made.
(c) Replication transparency: Allows duplicate resources in multiple components invis-

ibly. [16]
As it was mentioned earlier in this section, in distributed systems, components are presented
on different platforms and several components can cooperate with one another over the net-
work. In literature, there are several different architectures in distributed systems. In this paper
client-server model, multi-tier architecture and implementation of broker architecture CORBA
and the Service-Oriented Architecture (SOA) have been detailed.
1. Client-Server Model: This model or architecture is the most commonly used distributed
system architecture. This architecture classify devices into two sub-system. These are
client and server. In this model, server device have some resources or services which are
Client-Server Architecture Multi-Tier Architecture
Server
Database Server
reply reply
Database
Application
Server
request request
Client Client Client Client
Figure 2.5: Difference between Client-Server and Multi-Tier Architecture
ready to use by client [18]. Client sends request to the server to use these services provided
by server. Therefore, client needs to know information about servers to communicate. On
the other hand, server just provides requested service and does not need to know infor-
mation about client. This leads easy integration and scalability in distributed architecture.
2. Multi-Tier Architecture: This architecture is a type of client-server architecture. Differ-

ence between two architecture is the separation of different functionalities in the sys-
tem. For instance application processing, data management etc. are physically separated.
Therefore, it provides developers to easy integration and option of changing specific layer
without changing the entire application. It makes distributed systems more flexible and
reusable. Differences and similarities between these two architecture in structural level
can be seen at the figure 2.5.
3. Broker Architectural Style: This architecture style is for middleware systems used in dis-
tributed systems. It provides reliable communication between servers and clients. In
Figure 2.6: Implementation of CORBA
this architecture, a communication object called object request broker does communica-
tion. In this section, one of the implementation of broker architectural style called CORBA
(Common Object Request Broker Architecture) has been investigated. CORBA can be in-
vestigated in four main components. These are broker, stub, skeleton and bridge.
Broker coordinates communication between devices and handles the results and excep-
tions. “This can be either an invocation-oriented service, a document or message - ori-
ented broker to which clients send a message [18]” Moreover it organizes the service re-
quest, locates suitable servers and handles with requests and responses. Also it provides
APIs for clients to use the services and send and receive messages. In order to use as a
proxy for the client, there is a component called stubs, which is generated at the static
compilation time and deployed to the client. This client-side proxy components acts like
an arbiter between the client and the broker. Moreover it makes a remote object appears
like a local object in client [18].
Since there is a proxy in client side, there must be an another proxy in the server side.
This proxy in server side which is created by the service interface called skeleton. It is re-
sponsible for encapsulating low-level networking functions and providing high-level APIs
to negotiate between the server and the broker. Moreover it is responsible for handling
with communication and calling available service [18]. In addition to that, in order to
work with different brokers and different communication protocols such as DCOM, Java
Figure 2.7: Structure of SOA
CORBA brokers, there must a translator component. In CORBA architecture, it is called

bridge. These are optional components, which encapsulates the implementation details
when two brokers operate together and handles with communication and translates one
to another [18]. CORBA is an international standard and implementation of it in structural
level can be seen at the figure 2.6.
4. Service-Oriented Architecture: In this architecture, each service is a component of well-

defined, independent business functionality which can be used via a standard program-
ming interface. Connections among the services are provided by common and universal
message-oriented protocol. It is responsible for delivering requests and responses be-
tween services. This architecture is based on client-server architecture. However, it sup-
ports business-driven IT approach. That means an application consists of the software
services and its consumers [18]. Structure of this architecture can be seen at the figure
2.7.
2.3 Wireless Communication
In order to develop reliable operating system for Internet of Things devices, wireless communi-
cation must be investigated in detail. In this section, an overview has been given about wireless
communication and protocols. To better explanation and understanding, a conceptual model
Figure 2.8: OSI model [19]
was created called Open Systems Interconnection model (OSI). This is a conceptual model that
characterizes and standardise the communication functions without regard to their underlying
internal structure and technology. OSI model and explanation of each layer can be seen at the
figure 2.8.
To understand how wireless communication works, firstly physical layer and theory lies be-
hind it must be explained. The physical layer translates transmission and reception of wireless
waveforms and processes it. “It is a commonly acknowledged truth that the properties of the
transmission channel and the physical-layer shape significant parts of the protocol stack” [20].
Wireless communication fundamentals can be divided into three sub-section. These are fre-
Figure 2.9: Frequency Ranges [20]
quency allocation, modulation and demodulation and spread spectrum.
1. Frequency allocation: Every radio frequency based system has carrier frequency which
determines the propagation characteristics – for example, how well are obstacles like walls
penetrated – and the available capacity. A single frequency does not support any capacity.
Therefore frequency band which is a finite portion of the electromagnetic spectrum, is
used for communication purposes. In RF communication these frequencies starts from
Very Low Frequency(VLF) ends at Extremely High Frequency(EHF). Exact values of this
ranges can be seen at the figure 2.9 [20].
2. Modulation and Demodulation: In order to transmit data wirelessly, a digital signal must
be converted to a radio-, optical- or sonic signal that contains the same information as
the original signal. In digital electronics when systems communicate, they transfer digi-
tal data to each other which are the sequences of symbols and each symbol comes from
the channel alphabet are mapped to one a finite number of waveforms of the same finite
length. This process is called modulation. On the other hand, modulated signal must
be translated to meaningful data after it receives the signal. Demodulation is extracting
the original information-bearing signal from a modulated carrier wave. A demodulator
is an electronic circuit (or computer program in a software-defined radio) that is used to
recover the information content from the modulated carrier wave
There several types of modulation and demodulators. The signal output from a demod-
ulator may represent sound (an analog audio signal), images (an analog video signal) or
Figure 2.10: Types of Digital Modulation [21]
binary data (a digital signal).These terms are usually used in connection with radio re-
ceivers, but many other systems use different kinds of demodulators. For instance, in a
modem, which is a contraction of the terms modulator/demodulator, a demodulator is
used to extract a serial digital data stream from a carrier signal, which is used to carry it
through a telephone line, coaxial cable, or optical fiber.
In digital modulation, which is used in that particular project, there are three types. These
are Amplitude Shift Keying(ASK), Phase Shift Keying(PSK), Frequency Shift Keying (FSK).
The differences among them can be seen at the following figure 2.10.
3. Spread-Spectrum Communication: This is a dominating modulation and coding tech-

nique in wireless communication and networks. In this technique , the transmitted wave-
forms are way larger than needed to transmit the given data thus it occupies all the band-
width. The sender spreads the energy in the signal (baseband) over a wide frequency
band, for instance the bandwidth of the original signal is much narrower than that of the
transmitted signal. The receiver concentrates the energy in the broadband signal back
to the original baseband. Sender and receiver employ a special high frequency code se-
quence (key) to spread and concentrate the signal respectively. The code sequence is
denoted as pseudo-random-noise (PRN) and is only known by the sender and receiver.
For others, it will be experienced as white noise. The receiver restores the original signal
by synchronizing/correlating the incoming signal with the PRN sequence. In addition,
Figure 2.11: Spread Spectrum Communication [22]
the effects of narrowband noise and interference are reduced by using a wideband sig-
nal. Therefore, robustness of the system is increased and avoiding from multipath effect
can be achieved, though receiver for that technique is more complex than conventional
receivers [20]. It is illustrated at the figure 2.11.
There are two popular techniques in spread-spectrum communication
• Direct Sequence Spread Spectrum (DSSS)
– PRN code modulates the data signal directly (XOR)
– Phase shift modulation of baseband signal (BPSK, QPSK)
• Frequency Hopping Spread Spectrum (FHSS)
– PRN code basis for random hopping through a set of valid transmitter frequencies
– Frequency modulation of baseband signal (FSK)
So far, in this section theory behind the wireless communication has been detailed from
transceiver side to the receiver side. However, many devices are capable of communicating
wirelessly and without some regulations that would not be reliable communication type. There-
fore, some radio band (portions of radio spectrum) are reserved or defined internationally for
the use of radio frequency communication for industrial, scientific and medical applications.
These frequencies is called ISM radio bands. Example applications of these bands contain RF
process heating, microwave ovens and medical diathermy machines. Since these kind of devices
can create electromagnetic interference and disrupt radio communication in same frequencies,
these devices are limited to certain band in radio spectrum. Thus, communication devices must
handle and tolerate any interference caused by ISM applications. Moreover, particular restric-
tions exist with respect to transmission power and duty cycle for equipment that operate in
these frequency bands. The license-free radio bands ISM (Industrial, Scientific, Medical) make
up the basis for a range of different standards within wireless data communication. Most popu-
lar frequencies in this band are 315MHz, 433MHz, 868MHz, 2.4GHz, and 5.8GHz.Radio modems
for simple digital wireless point-to-point communication in the ISM bands are readily available
and quite easy to use (relatively low data rates). Typical applications: cable replacements, re-
mote control, access control, telemetry.
There exists an almost bewildering choice of connectivity options for electronics engineers
and application developers working on products and systems for the Internet of Things (IoT).
Depending on the application, factors such as range, data requirements, security and power
demands and battery life will dictate the choice of one or some form of combination of tech-
nologies. In this section, only the major communication protocols/technologies in embedded
systems have been mentioned.
1. Bluetooth: For at least fifteen years in computing and many consumer products in the
market, Bluetooth has become quite popular and important short-range communication
protocol. As especially for wearable products is the key technology because of its spec-
ification. In IOT devices, it is used via a smartphone in most of the application. More-
over, it is expected to increase its popularity and usage in IOT applications with the new
Bluetooth low-energy (also known as Bluetooth Smart) protocol. The new protocol offers
similar range to Bluetooth and significantly reduced power consumption thus it suits well
to embedded devices and applications [23].
• Standard: Bluetooth 4.2 core specification
• Frequency: 2.4GHz (ISM)
• Range: 50-150m (Smart/BLE)
• Data Rates: 1Mbps (Smart/BLE)
2. Zigbee: Widely used protocol especially in industrial application. It has different profiles
for different applications. Particularly ZigBeePRO and ZigBee Remote Control protocols
are based on the IEEE 802.15.4 protocol (its and industrial standard for wireless technol-
ogy), therefore, it is suitable for industrial applications which need infrequent data ex-
changes at low data-rates. In addition to that, it offers 100m range that makes it more
suitable for home or building applications [23].
• Standard: ZigBee 3.0 based on IEEE802.15.4
• Frequency: 2.4GHz
• Range: 10-100m
• Data Rates: 250kbps
3. Wi-Fi: Wi-Fi is the leading technology in electronics and computer systems after the pro-
liferation of the internet. However, it used to be used in personal computing rather than
embedded or electronics application. It requires bigger stack and well-built infrastruc-
ture. The technological developments in IC design and electronics, led to more process-
ing power with bigger memory for the same price. Therefore, it has become a reliable and
easy solution for embedded applications as well. It offers fast and reliable data transfer
and the ability to handle high volume of data [23].
• Standard: Based on 802.11n (most common usage in homes today)
• Frequencies: 2.4GHz and 5GHz bands
• Range: Approximately 50m
• Data Rates: 600 Mbps maximum, but 150-200Mbps is more typical, depending on
channel frequency used and number of antennas (latest 802.11-ac standard should
offer 500Mbps to 1Gbps)
4. NFC: Also known as Near Field Communication enables very simple communication be-
tween two electronic devices with safe two-way interaction. Especially along with the digi-
tal revolution, it has started to find wide area of usage in contactless payment transactions,
accessing digital content and connecting electronic devices. Mainly, it has increased the
capability and quality of contactless card technology with providing secure communica-
tion at a distance that is less than 4cm [23].
Figure 2.12: Overview of Wireless Technologies in the market [24]
• Standard: ISO/IEC 18000-3
• Frequency: 13.56MHz (ISM)
• Range: 10cm
• Data Rates: 100–420kbps
Overview of most of the short-range wireless technology popular in nowadays can be seen at
the figure 2.12.
2.4 Related Work
In order to develop an efficient and novel design in operating system, related works and em-
bedded operating systems which are widely used currently in industry and academy must be
investigated in detail. Therefore, specification about particular operating system must be de-
cided according to market needs, current market situation and related work or similar products.
In this section six operating system that are inspired us to develop such an operating system
have been mentioned. These are most popular operating system currently both in the industry
and academy.
1. FreeRTOS: It is a real time operating system as it can be seen from its name and specifi-
cally designed for embedded devices. It is an open source operating system in addition, it
provides compiler toolchains and has support for many different architectures. Designing
purpose of this operation system is to be “small, simple and easy to use” [25].
Same as all operating systems, main task of FreeRTOS is to run tasks reliably. Therefore,
most of the code is related with scheduling, prioritizing and running user-specific tasks.
Since the simplicity of its design, the kernel contains only three or four C files , thus it
makes the all system portable and easy to maintain. Although its simplicity, multiple
threads or tasks, mutexes, semaphores and software timers are provided for users. More-
over it provides four memory allocation schemes . These are allocate only, allocate and
free with a very simple, fast, algorithm, a more complex but fast allocate and free algo-
rithm with memory coalescence and C library allocate and free with some mutual exclu-
sion protection [26].
Operating systems like Linux or Microsoft Windows, more features like device drivers, ad-
vanced memory management, user accounts, and networking can be found. However,
the emphasis of FreeRTOS is on compactness and speed execution thus, it can be seen as
a thread library rather than operating system although it provides command line interface
and POSIX-like I/O abstraction add-ons.
2. Contiki-OS: Contiki-OS firstly introduced to the market in 2003. Like FreeRTOS, it has
widely usage in resource-limited systems. It requires 10kb of Ram and 30kb of Rom to
function and it defines itself as an ultimate OS for network devices. Contiki mainly de-
signed to connect low-power microcontrollers to the Internet and Wide-Area Networks.
It provides three popular network mechanism and these are the uIP TCP/IP stack, which
provides IPv4 networking, the uIPv6 stack, which provides IPv6 networking, and the Rime
stack, which is a set of custom lightweight networking protocols designed specifically for
low-power wireless networks. The IPv6 stack was contributed by Cisco and was, at the
time of release, the smallest IPv6 stack to receive the IPv6 Ready certification [27].
Furthermore, to achieve multitasking in embedded systems, programming model of Con-

tiki is based on protothreads. A protothread is a memory-efficient programming abstrac-
tion that shares features of both multi-threading and event-driven programming to attain
a low memory overhead of each protothread [28].
3. Micro C/OS: “It is a public domain real-time operating system which is representative of
current commercial RTOSes which employ static priority based scheduling algorithms”
[29]. It supports 64 different priority level for user-defined functions. Main purposes of
the design are determinism, modularity and scalability. MicroC/OS allows defining sev-
eral functions in C, each of which can execute as an independent thread or task. Each
task runs at a different priority, and runs as if it owns the CPU. Lower priority tasks can be
pre-empted by higher priority tasks at any time. Higher priority tasks use operating sys-
tem services (such as a delay or event) to allow lower priority tasks to execute. Moreover,
it provides standard operating services such as IPC mechanism, task management, mem-
ory management and timing. It has two version called pre-emptive and non-preemptive
version.
4. Tiny OS: One of the oldest open source operating system is TinyOS. It is the inspiration
for all the other operating system after it. It is the first embedded operating system that
is designed for low-power embedded devices and wireless sensor networks. It introduce a
term called mote which means a microcontroller based node in wireless sensor network
and , capable of reading sensory information, processing and exchanging its data with
other nodes [30]. TinyOS applications are developed in the programming language nesC,
a dialect of the C language optimized for the memory limits of sensor networks. The ex-
tra tools of the system are mainly in the form of Java and shell script front-ends. Every
program are built as components and those components communicates among them via
interfaces provided by operating systems. Moreover, TinyOS provides interfaces and com-
ponents for common abstractions such as packet communication, routing, sensing, actu-
ation and storage. All the I/O operations that last longer than a few hundred microseconds
are asynchronous and have a call-back since the systems is full-non blocking and has only
one call stack. To enable the native compiler to better optimize across call boundaries,
TinyOS uses nesC’s features to link these call-backs, called events, statically. While being
non-blocking enables TinyOS to maintain high concurrency with one stack, it forces pro-
grammers to write complex logic by using many small event handlers. To support larger
computations, TinyOS provides tasks, which are similar to a Deferred Procedure Call and
interrupt handler bottom halves. A TinyOS component can post a task, which the operat-
ing system will schedule to run later. Tasks are non-pre-emptive and run in first in, first
out order. This simple concurrency model is typically sufficient for I/O based applica-
tions, but its difficulty with CPU-heavy applications has led to developing a thread library
for the particular system, called TOSThreads.
5. QNX: This operating system is one of the most popular commercial operating system in
the market. It is also most traditional Real-Time Operating System. It is based on the
micro-kernel architecture and it supports POSIX standards. In this micro-kernel imple-
mentation as it is discussed earlier in this paper, only four operating system services are
implemented and these are task scheduling, inter task communication, low level network
communication and interrupt handling. All the remaining parts of the system (device
drivers included) are implemented as user tasks, making the kernel fast, reliable and small
[31]. All this kernel services known as servers in QNX. Therefore, it is different from mono-
lithic kernel. The operating system gives user to turn off and on any operating system ser-
vice without needing any change in the operating system itself. QNX Neutrino has been
ported to a number of platforms and now runs on practically any modern CPU that is used
in the embedded market. QNX neutrino supports ARM, MIPS, PowerPC, SH4 and the PC
architecture.
6. Embedded Linux(Yocto Project) : In recent years, Linux has started to become popular
on embedded applications and devices especially consumer gadgets, telecom routers and
switches, Internet appliances and automotive applications. “Because of the modular na-
ture of Linux, it is easy to slim down the operating environment by removing utility pro-
grams, tools, and other system services that are not needed in an embedded environment
“ [32]. The Yocto Project has the aim and objective of attempting to improve the lives of de-
Embedded
FreeRTOS Contiki OS Micro C/OS TinyOS QNX Linux(Yocto
Project)
Relatively
Big
Relatively Small(20/
Small(5kb Small(5kb size(however
Footprint Small(10kb 30kb to Big Size
to 10kb) to 24kb) it is config-
to 30kb) several
urable)
hundred
Kernel
Event/driven Linux Ker-
architec- Microkernel Microkernel Monolithic Microkernel
kernel nel(monolithic)
ture
Intel 8088, ARM,32
MSP430 x86, MIPS, and 64bit
More than More than family, PowerPC, x86,PowerPC,
30 different AVR,MSP430, 30 different Atmega128 SH-4, MIPS(all
Platforms
architec- ARM7, PIC32 architec- family, and ARM, major em-
ture ture the Intel Stron- bedded
px27ax gARM, architec-
XScale tures)
Table 2.1: Evaluation of Embedded Operating Systems
velopers of customised Linux systems supporting the ARM, MIPS, PowerPC and x86/x86
64 architectures. A key part of this is an open source build system, based around the
Open-Embedded architecture, which enables developers to create their own Linux distri-
bution specific to their environment. This reference implementation of Open-Embedded
is called Poky. As well as building Linux systems, there is also an ability to generate a
toolchain for cross compilation and a Software Development Kit (SDK) tailored to their
own distribution, also referred to as the Application Developer Toolkit (ADT). The project
tries to be software and vendor agnostic.
Detailed comparison and evaluation about these six operating systems is given in the tables
2.1 and 2.2.
Positive Points Negative Points Network or/and IOT feature
 Open Source
 Large Support and  Overhead may occur in  No specific support for
Community particular configurations. Networking or IOT
Free RTOS  Optional devices
preemptive  The instruction documents
Scheduling and are not free.
other scheduling
algorithms
 Device Support
 Optional  Not too many  Network Simulator

preemptive architectures(especially arm
Scheduling cortex m series)  Sensor/Network Specific
Contiki OS  GUI OS
 Downloading code  Relatively Poor
at run time documentation  GUI
 Light weight
 Embedded System
Tools
 Allow stack growth  Supports networking
Micro C/OS of tasks to be  Expensive License technology IPV4
monitored
 Preemptive
multitasking real-
time kernel
 Limited Physical
 Flexibility  Lack preemptive real-time Parallelism and
scheduling Controller Hierarchy
TinyOS  Low power  Lack of flexibility  Specifically developed for
wireless sensor
 Small footprint  Lack of Virtual Memory applications
 Can run Linux

applications with no  Expensive Licence  Supports networking
modifications technologies include
 Cross-compile IPv4, IPv6, IPSec, FTP,
advantage HTTP, SSH and Telnet.
QNX  Has a solid, stable
API, which is very  Limited choice of hardware  Has configuration for
well documented. platform Resource Sharing
 QNX has only one
graphical toolkit
supplied (Photon)
Embedded  Linux Kernel

Linux(Yocto  Big size Linux kernel  Supports networking
Project)  Big community and technology IPV4
good support
Table 2.2: Comparison of Operating Systems Embedded

Chapter 3
Details and Features
After detail research about the operating systems in the market, the features and the details
about particular operating systems have been decided. In addition, the needs of industry and
past works in academy have been considered. The features about this project are mentioned in
earlier sections. However, after consideration of time and workload some changes in features
must have been done to achieve the implementation of all the features. Therefore, implemen-
tation of different power modes and security layer had been changed as possible future works.
The particular operating system provides only one power mode that is designed to be low-power
mode. In addition to that, instead of implementing completely new security layer, a new com-
munication protocol for IOT service has been designed and implemented.
In this particular section, you will find more information about details and features. Firstly,
designing steps of how to achieve small footprint and efficient task execution are mentioned. In
addition, information about reduced-size operating system services and real-time compatible
priority-driven scheduler can be found. Later on, our novel design for Kernel architecture is
detailed and how it makes operating system more user-friendly is explained. At the end of this
chapter, the proposed IOT service and how it enables resource sharing between devices can be
found.
35
CHAPTER 3. DETAILS AND FEATURES 36
3.1 Simple and Light-weight Design
As it was discussed earlier in this paper, technological development in silicon industry does not
have same effects on CPU power and memory sizes. Since the beginning of 80s, CPU speed of
the microcontroller has been increasing more than memory size. With the help of new powerful
inside of embedded systems, developers demand more task from a microcontroller. This leads
complicated and big operating system and libraries. Therefore, most of the operating system
have started to suffer from lack of memory.
The term of embedded devices can also be defined as memory-constrained or resource-
constrained systems. It is because usually at the best case microcontrollers have a few hundred
MHz of CPU and few hundreds of ram and flash. Therefore, an embedded operating system
must be small footprint in order to work efficiently in embedded devices. In addition to that,
there must be enough space in the ram and flash for run-time and user-defined applications.
This project is based on the idea of simplicity and smallest size that can be achieved. Com-
plicated operating systems require lots of memory space. Although memory requirements can
be satisfied buying external flash memory or selecting bigger microcontroller, it is very expen-
sive. Since the concept of IOT is spreading around our daily life, this kind of cost could get very
high in mass production. However, simple and efficient operating system can be achieved if
this operating specifically developed and optimized for IOT concept. In order to develop such
an operating system, the needs from an operating system must be known and system compo-
nents must be defined well.
Firstly, kernel function of the operating system was divided into two sections. These are
portable kernel and native kernel libraries. Moreover, portable kernel also can be divided into
three section. Firstly, a component called dispatcher in the particular operating systems are
written in assembly language. Dispatcher is the part of kernel that is triggered by an inter-
rupt and receives the full register set of the program that was running at the time of interrupt
(with the exception of the program counter, which is presumably propagated through a mutu-
ally agreed upon ’volatile’ register or some such). Thus, the dispatcher must be carefully written
to store the current state of register banks as its first operation upon being triggered. In short,
the dispatcher itself has no immediate context and thus does not suffer from the same problem.
Native Kernel
Mutex/Semaphore Library
Memory Allocation/
Task/Process Library Scheduler
Stack Functions
Portable Kernel
Sysick Library Dispatcher Vendor Specific Files
Figure 3.1: Kernel Structure
The second part of portable kernel which can trigger dispatcher when it is necessary, is system
tick. A tick is an arbitrary unit for measuring internal system time. There is usually an oper-
ating system internal counter for ticks; the current time and date used by various functions of
the operating system are derived from that counter. How many milliseconds a tick represents
depends on the operating system, and may even vary between installations. Use the OS’s mech-
anisms to convert ticks into seconds. The third part of the portable kernel is interrupt handlers
for exceptions such as system-level service and supervisor call. This two function handles the
registers during context switch or exception handling. Detailed information about native kernel
and portable kernel will be mentioned in section four.
As it was mentioned in the paragraph below, portable kernel has some devices specific defi-
nition and functions and also device definition files provided by vendor of particular device(processor
or microcontroller). Therefore, since this operating system is designed to support many differ-
ent processor and architecture, this part was separated from native kernel to make operating
system more portable.
First optimization in particular operating system was made in kernel side. All the compo-
nents of the kernel can be seen at the figure 3.1. Most of the embedded operating system
contains a large kernel which has more capabilities. For example, usually except scheduler
and dispatcher, they support operating system event to provide privileges to operating system
services. In addition to that, they provide software timer for upper layer of the system. Com-
plicated mutex and semaphore infrastructure is also common component of other operating
system. Furthermore, memory protection layer is implanted for the memory management for
more complicated CPUs. However, most of the embedded system contains low-power and small
size processor. In this particular architecture, as it can be seen, kernel tried to develop as small
as possible. Major task of kernel is to achieve reliable execution of task and provide required
infrastructure for semaphores and mutexes. APIs and other service of the operating system can
handle other extra features and these can be configured from a file. Therefore, even for very
basic embedded application, this operating system can be implemented with smaller footprint.
However, optimization in kernel is not enough to reduce to size of the operating system.
Generally, operating system services require lots of memory and since most of them are exe-
cuted periodically, require CPU power as well. An Operating System provides services to both
the users and to the programs. Common services provided by operating systems are program
execution, I/O operations, file system manipulation, communication, error detection, resource
allocation, protection [10]. However, in embedded systems and application most of them are
implemented in vain. In embedded applications, user must decide I/O operation, since devices
are application and I/O specific. Memory allocation can be done calling kernel functions, there
is no need to have a services running for it. Moreover, usually file system is not a necessity in
embedded applications due to memory restriction. Therefore, implementing service for that
would be a pointless act. In this project, needs of a typical embedded application were inves-
tigated in detail. As a result, operating system services reduced to two services. One of them is
booting service, which just is executed only once at the beginning to set and configure required
hardware. The other one is idle service, which puts operating system to sleep state, if there is
no task/process to schedule. User can be controlled all the other operating system services by
using APIs for specific part of the system. Moreover, there is IOT service that is implemented on
top of everything in the system and it has the higher priority among the task. However, since
it is not implemented inside of operating system service component and has different concept
than traditional operating system services, it does not count as operating system service in this
Application Layer (User-defined functions)
API
OS Services
Booting Service Idle function
Figure 3.2: Structure of OS Services with API and Application Layer
particular operating system. Details about operating system services and API can be found later
on this paper. In addition, structure of operating system services with API and application layer
can be seen at the figure 3.2.
3.2 Two Dimensional Kernel Architecture
As it was mentioned before, one of the feature of proposed operating system is a novel and sim-
ple architecture. Operating system has been divided into seven components. From bottom to
top these are hardware driver, hardware abstraction layer, scheduler and dispatcher, operating
system services, application layer, communication layer and IOT service layer. Proposed archi-
tecture of the embedded operating system can be seen at the figure 3.3.
There are currently many types of kernel architecture for operating systems in the market
and academy. As it was discussed before, two of must popular ones are monolithic and mi-
crokernel architecture. However, for instance also hybrid kernel architecture is widely used.
Especially systems based on Microsoft Windows has this architecture. They are similar to micro
kernels, except they include some additional code in kernel-space to increase the performance.
IoT Service & Application
Security Layer
Commuication Layer Application Layer
OS Services
Scheduler, Dispatcher and HAL
HARDWARE
Figure 3.3: Proposed Operating System Architecture
These kernels represent a compromise that was implemented by some developers before it was
demonstrated that pure micro kernels could provide high performance. These types of kernels
are extensions of micro kernels with some properties of monolithic kernels. In addition, they are
similar to micro kernels, except they include some additional code in kernel-space to increase
performance. In addition, there is nanokernel, which delegates virtually all services (including
even the most basic ones like interrupt controllers or the timer) to device drivers to make the
kernel memory requirement even smaller than a traditional microkernel. Moreover exokernels
which is very different from other traditional kernels. Exokernels are extremely small. However,
they are accompanied by library operating systems providing application developers with the
functionalities of a conventional operating system.
As it can be seen, there is no need to have a different kernel architecture unless it is based on
a novel idea. In this project, a kernel architecture is proposed as a solution that cannot be solved
using other kernels. Firstly, our embedded operating system is based on simplicity and small
size. All of those kernel architectures are not specifically developed for small and low power mi-
crocontroller. Since the particular operating system is developed specifically for this purpose,
most of the operating system parts can be configurable to reduce to size for different applica-
tions. Basic configuration of the operating system provides multitasking, memory allocation
and real-time task execution. Since even basic configuration has the same basic capabilities
with other kernel, this architecture is called microscopic kernel architecture.
Second purpose to develop new architecture is to achieve user-friendly and secure operat-
ing systems. Most of the embedded developers tend to develop basic C functions and libraries
for the specific application. Usually using an operating system for this kind of application it is
not the first choice of the developer due to complexity of software and necessity of learning new
software. Therefore, idea behind this kernel architecture is to create a user space (application
layer) for developer. Developers can control hardware like they did before, and they do not need
to develop or learn new software to implement wireless networks or deal with other communi-
cations in the systems. That is the first dimension of our proposed architecture. User dimension
just need to develop their code like there is no operating system. This can be seen from the figure
above, our first dimension is start from right hand side until the communication abstraction. In
addition, also developers who want to contribute or change some kernel functions according to
their need in application, there is another dimension from bottom the top called supervisor di-
mension in the structure of the operating system. Therefore, this architecture actually called as
microscopic two-dimensional kernel. With the help of two dimension, implementation of the
operating system is easier. Moreover, development of operating system for future work became
easier and more collaborative. Details about kernel operation and functionality can be found
later on this paper.
3.3 Internet of Things Service and Resource Sharing
Last feature and purpose of the particular project is to achieve reliable resource and information
sharing. Earlier in this paper, six different operating systems for embedded devices have been
mentioned. Moreover, during the development of the project, many other operating systems
were investigated. Although most of them have some feature or libraries specifically provided
for Internet of things, none of them has an OS service for this kind of applications.
Our main idea is to develop an embedded operating system not only for internet of things
IoT Service Layer
Distributed System
Middleware like Layer
Radio Protocol
IOT Service API
Stack
Radio Module
Application
Layer
Communication
Layer
Figure 3.4: Structure of IOT Service
devices but also for all kind of wireless network. Large networks which may be formed by hun-
dreds or thousands of nodes, can be created easily. However, in the application side, this new
trend brought some new problem to the networks and software. Since wireless communica-
tion not reliable as wired communication, there are always some minor errors such as missing
packet and etc. in the wireless network. In addition to that, most of the sensor network are used
in critical measurements in harsh environments. Sensor nodes may become unavailable be-
cause of lots of different reasons. Sometimes this failure may not create big problems. However
sometimes these kind of failures may cause of shutting down whole network. To overcome these
problems, different kind of approaches can be used. Most of these approaches requires reliable
communication and information sharing. Therefore, in this particular project, a novel operat-
ing system called Internet of Things service has been introduced as an infrastructure. Structure
of proposed architecture can be seen at the figure 3.4.
Earlier period during this project, Bluetooth protocols was the base of this service. The IOT
service was implemented on top Bluetooth smart protocol. However, Bluetooth stack has a large
footprint. Since our focus is to develop, a very small size embedded operating system that was
not an optimal solution. Bluetooth stack was tried to be reduced but even optimized stack for
specific application was too big. Therefore, a new radio library and a specific radio protocol for
wireless sensor networks and Internet of Things decided to develop. Lightweight protocol stack
for this particular protocol is implemented into IOT service.
Like every other operating system service, IOT service periodically does particular task to
share data between other devices in the network. Actually, this IOT service can be considered
as distributed system middleware. While developing this services, distributed system concepts
such as CORBA and SOA architecture is used as conceptual base. In addition to that, a basic
mail architecture has been added to this service. Therefore, this service can be seen as hybrid
of all these three architecture. On the other hand, this service handles all wireless communica-
tion and in the meantime, it does not affect the execution of other user applications or tasks.
User can interact with the service using IOT service API. Moreover two device types have been
introduced for the network structure, thus, user needs to select device type before the develop-
ment. However as it was mentioned before, user does not need to develop and library to enable
resource or information sharing. More information about Internet of Things Services and radio
library can be found later on this paper.
3.4 3.4. Development Platform
In this project, a small embedded operating system is developed for low-cost and small (memory-
constrained) microcontrollers. Although this project aim to support all this kind of microcon-
trollers, for development one platform must have been selected. Therefore, a system-on chip
from Nordic Semiconductor called nRF51822 was chosen for this particular project.
This integrated circuit is a flexible and powerful system on chip, which contains a microcon-
troller and radio module ideally suited for Bluetooth applications. The nrf518221 is based on
32-bit ARM cortex M0 CPU with 256kB/128kB flash + 32kB/16kB RAM for improved application
performance. In addition, embedded transceiver works at 2.4 GHz and supports both Bluetooth
Smart and the Nordic Gazell 2.4 GHz protocol stack [33].
The SOC contains a rich option for analog and digital peripherals that can be executed with-
out interrupting CPU through Programmable Peripheral Interconnect system. “A flexible 31-pin
GPIO mapping scheme allows I/O like serial interfaces, PWM and quadrature demodulator to
be mapped to any device pin as dictated by PCB requirements” [33]. Therefore, this leads to
have more flexible designs with pin-out location and function.
For CPU architecture, ARM offers large variety of processor architecture for different types
of embedded application. As it was mentioned before, in this SOC contains ARM cortex M0
CPU with 16-bit instruction set with 32-bit extension (Thumb-2 technology) which provides a
small-memory-footprint in high-density code. In addition to that, the processor uses a single-
cycle 32-bit multiplier, a 3-stage pipeline, and a Nested Vector Interrupt Controller (NVIC) to
achieve simple and efficient program execution [34]. Furthermore, ARM provides a hardware
abstraction layer called The ARM Cortex Microcontroller Software Interface Standard (CMSIS)
to achieve further compatibility with other architecture such as ARM Cortex M3 based devices
[34].
“The nRF51 series power management system is orthogonal and highly flexible with only
simple ON or OFF modes governing a whole device.” [35]. When it is set to System OFF mode,
sections in the RAM can be retained although everything is powered down. System can be easily
set to System ON mode through reset or it can be waken up from all GPIOs. Once system is on,
all functional blocks can be used again and switch to the IDLE mode [35].
The two pin Serial Wire Debug interface (provided as a part of the Debug Access Port, DAP)
offers a flexible and powerful mechanism for non-intrusive program code debugging. This in-
cludes adding breakpoints in the code and performing single stepping [35]. The block diagram
of the SOC can be seen at the figure 3.5.
The reasons of the choice of particular processor architecture as follows;
• The smallest ARM processor: Right now in the market, it is the smallest ARM architecture
can ever be found. Not only code density and energy efficiency but also forward compat-
ibility for other bigger architecture such as Cortex-M3 and Cortex-M4 processors makes
this architecture more suitable for this project.
• Low power: The Cortex-M0 processor, which consumes as little as 12.5µW/MHz (90LP
process, minimal configuration) in an area of under 12 K gates is the leader microcon-
troller in the market now.
• Simplicity: With just 56 instructions, it is possible to master quickly the entire Cortex-
Figure 3.5: Block Diagram of NRF51822 [34]

M0 instruction and its C friendly architecture, making development simple and fast. The
option for fully deterministic instruction and interrupt timing makes it easy to calculate
response times.
• Optimized connectivity: Most important feature of this architecture is to be optimized for

connectivity. It is specifically design to be used in application with wireless protocols such
as Bluetooth Low Energy (BLE), IEEE 802.15 and Z-wave [36].
On the other hand, other components of SOC that is radio transceiver is worth to mention as
well. Nordic Semiconductor designed and optimized this transceiver specifically to operate in
the worldwide ISM frequency band at 2.400 to 2.4835 GHz. Moreover, it its highly configurable
transceiever, thus radio modulation modes and packet structures can be configured to support
large variety of protocols such as Bluetooth low energy, ANT, Enhanced ShockBurst [34].
The transceiver receives and transmits data directly to and from system memory for flexible
and efficient packet data management. The on chip transceiver has the following features:
• General modulation features
– GFSK modulation
– Data whitening
– On-air data rates
* 250 kbps
* 1 Mbps
* 2 Mbps
• Transmitter with programmable output power of +4 dBm to -20 dBm, in 4 dB steps
• Transmitter whisper mode -30 dBm
• RSSI function (1 dB resolution)
• Receiver with integrated channel filters achieving maximum sensitivity
– -96 dBm at 250 kbps

Figure 3.6: Nrf51822 Radio Module Block Diagram [35]
– -93 dBm at 1 Mbps BLE
– -90 dBm at 1 Mbps
– -85 dBm at 2 Mbps
• RF Synthesizer
– 1 MHz frequency programming resolution
– 1 MHz non-overlapping channel spacing at 1 Mbps and 250 kbps
– 2 MHz non-overlapping channel spacing at 2 Mbps
– Works with low-cost ± 60 ppm 16 MHz crystal oscillators
• Baseband controller
– EasyDMA RX and TX packet transfer directly to and from RAM
– Dynamic payload length
– On-the-fly packet assembly/disassembly and AES CCM payload encryption
– 8 bit, 16 bit, and 24 bit CRC check (programmable polynomial and initial value) [34]
The block diagram of on chip transceiver can be seen at figure 3.6 .

Chapter 4
Hardware Abstraction Layer and Drivers
In order to use the capabilities of the microcontroller peripherals, a piece of software which is
specifically developed for this purpose, what an operating system needs. Driver provides a soft-
ware interface to hardware devices, enabling operating systems and other programs to access
hardware functions without requiring details of the hardware being used.
A driver communicates with the device through the computer bus or communications sub-
system to which the hardware connects. When a calling program invokes a routine in the driver,
the driver issues commands to the device. Once the device sends data back to the driver, the
driver may invoke routines in the original calling program. Drivers are hardware dependent
and operating-system-specific. They usually provide the interrupt handling required for any
necessary asynchronous time-dependent hardware interface. As it was mentioned before in
this particular project, nRF51822 SOC has been used as target platform. Therefore, all the hard-
ware drivers in this project are written for that microcontroller. Since this SOC contains a mi-
crocontroller, also it comes with lots of peripherals to use. Some of them are implemented in
this particular project and the others can be developed later on and integrated easily into the
operating system.
In this project, two type of development board which contains particular SOC have been
used to test and verify the operating system. Thus, firstly board specific libraries was developed.
The purpose of these libraries are to create common ground for development between boards.
Since there are two type of development board, all the pins for LED and button connection,
UART pins for J-link connection are different. All the libraries are implemented on top of board
48
CHAPTER 4. HARDWARE ABSTRACTION LAYER AND DRIVERS 49
GPIO library Busy-wait library Timer library Radio Library UART Library
Board Specific/Abstraction Library
Hardware
Figure 4.1: Structure of Hardware Drivers
libraries. Therefore, there is no need to develop or use two different driver libraries for two
different boards. Structure of hardware drivers can be seen at the figure 4.1 .
One of the most important library in hardware drivers for most of the embedded operating
system is general-purpose input-output libraries or widely known as GPIO. This library is used
to communicate or interact with other components, actuators or devices. Therefore, even to
light a small LED, this library must be used. To reduce the size of the GPIO library, only basic
functionality is implemented. Most of the cases, main tasks of this library set particular pins
as input or output, set output pin values to logic high or low and configure pull-up and pull-
down of pins. Therefore, only three functions are developed and implemented. Firstly, configure
function needs three argument. These are pin number, direction and pull-up/down option.
User can configure desired pin according to needs. Pin-set function sets desired pin to logic
high and pin-clear function sets desired pin to logic low.
Timer functionality of device is a crucial for many applications but especially for real-time
applications. In this particular SOC, there are three timers. One of the can be configured up to
32bit and other tow can be configured up to 16bit. Designing of the timer library is based on very
simple architecture. Initialize, start and stop functions are implemented according to nature of
timers. User can set interval and the address of call-back functions. In computer programming,
a call-back is a piece of executable code that is passed as an argument to other code, which is
expected to call back the argument at some convenient time. The invocation may be immediate
as in a synchronous call-back, or it might happen at later time as in an asynchronous call-back.
In all cases, the intention is to specify a function or subroutine as an entity that is, depending
on the language, more or less similar to a variable. In this library call-back functions provides
handling with timer interrupt to user without dealing with the timer registers. In addition, also
the SOC contains a real time counter. This timer has been used as SYSTICK timer. This library
cannot be used for other purposes or changed by user and is implemented into portable kernel.
Busy-waiting, busy-looping or spinning is a technique in which a process repeatedly checks
to see if a condition is true, such as whether keyboard input or a lock is available. Bust-wait can
also be used to generate an arbitrary time delay, a technique that was necessary on systems that
lacked a method of waiting a specific length of time. Processor speeds vary from computer to
computer, especially as some processors are designed to dynamically adjust speed based on ex-
ternal factors, such as the load on the operating system. As such, spinning as a time delay tech-
nique often produces unpredictable or even inconsistent results unless code is implemented
to determine how quickly the processor can execute a "do nothing" loop, or the looping code
explicitly checks a real-time clock. To create a common wait/delay function for all the appli-
cations, bust-wait library is implemented in this level. Basically user can delay execution for
desired time. In addition, with the help of two functions, this time can be set in microseconds
or milliseconds precision. These functions executes no operation instruction of the processor
several time to reach one microsecond and then using a loop, this delay is extended. While
busy-wait functions are being executed, the operating system keeps handling with interrupts
and other operating system services.
A universal asynchronous receiver/transmitter, abbreviated UART is a computer hardware
device that translates data between parallel and serial forms. UARTs are commonly used in con-
junction with communication standards such as TIA (formerly EIA) RS-232, RS-422 or RS-485.
The universal designation indicates that the data format and transmission speeds are config-
urable. The electric signalling levels and methods (such as differential signaling etc.) are han-
dled by a driver circuit external to the UART. A UART is usually an individual (or part of an)
integrated circuit (IC) used for serial communications over a computer or peripheral device se-
rial port. UARTs are now commonly included in microcontrollers. In this particular project
UART library is implemented for the debugging issues in real time applications. Simple UART
library is formed by only three functions. UART -configure function configures communication
for desired baud-rate and sets corresponding bits to prepare physical layer of communication
correctly. UART -send function needs avoid pointer, and length of the UART library to send.
UART -Receive function works and needs same arguments. In order to keep UART library sim-
ple, none of other feature or interrupt of UART module were not implemented in this particular
project.
Wireless communication particularly most important part of the project. In order to achieve
relible communication between devices, library must be developed carefully. Moreover that
since a small protocol has been created for this project, library for radio module must be devel-
oped accordingly the stack of the protocol. Mainly this library provides some functionality as
other communication libraries. Configuration, send and receive function works like the func-
tions in UART library. However, radio library has more features, handles with interrupts, and
supports other functionalities. Details about this library is given in the section called “Radio Li-
brary, Stack and Protocol” due to make this parts of paper more understandable. Choice of the
implement other modules and peripherals inside of the microcontroller has left to user. Since
for the test scenarios implementation of particular peripherals was enough, libraries such as
I2C, SPI or ADC were not implemented.
Another part of the system which will be explaining rest of the section is hardware abstrac-
tion layer. Hardware abstraction layer also known as HAL is formed by routines and functions in
the operating system which emulates hardware and platform specific functionalities and pro-
vides direct access layer to hardware resources for the application layer.
At the past, the computer or embedded systems did not have any kind of software layer,
which works as the hardware abstraction layer. Therefore, developers working with those de-
vices would have to know how each hardware device communicated with the rest of the sys-
tem. This was very challenging for the software developers since every part of the hardware in
the system had to be known to ensure the software’s compatibility. With HAL, rather than the
program communicating directly with the hardware device, it communicates to the operating
system what the device should do, which then generates a hardware-dependent instruction to
the device. For that reason, the software started to be compatible with any device without need
for the information and details about the operation of hardware.

An example of this might be a "Joystick" abstraction. The joystick device, of which there are
many physical implementations, is readable/writable through an API which many joystick-like
devices might share. Most joystick-devices might report movement directions. Many joystick-
devices might have sensitivity-settings that can be configured by an outside application. A Joy-
stick abstraction hides details (e.g., register formats, I2C address) of the hardware so that a pro-
grammer using the abstracted API needn’t understand the details of the device’s physical inter-
face. This also allows code reuse since the same code can process standardized messages from
any kind of implementation which supplies the "joystick" abstraction. A "nudge forward" can
be from a potentiometer or from a capacitive touch sensor that recognizes "swipe" gestures, as
long as they both provide a signal related to "movement".
In this embedded operating system, hardware abstraction layer used as a very thin layer
that is responsible from translation of hardware drivers. Upper layers of operating system and
hardware drivers can be seen as two different group of people who speak different languages.
HAL acts like a translator between two layers and translates driver function into easier functions.
Basically HAL is divided into seven basic type of drivers which can be found in microcontrollers.
Translation is done using specific structures for all types of device.
typedef s t r u c t {
void ( * configure ) ( int , hal_gpio_dir_t , hal _gpio _pu ll _t ) ;
void ( * s e t ) ( uint8_t ) ;
void ( * c l e a r ) ( uint8_t ) ;
} hal_gpio_dev_t ;
s t a t i c hal_gpio_dev_t dev_gpio = {
&gpio_pin_configure ,
&gpio_pin_set ,
&gpio_pin_clear
};
Listing 4.1: Hal structure for GPIO
As it can be seen from the listing 4.1, GPIO structure has been designed accordingly to driver.
Then device defined as static global structure and drivers function are assigned. With the help
of that, device can be used by user easily. Moreover it increases readability of the code.
void ( * open ) ( void ) ;
void ( * configure ) ( void * ) ;
void ( * send ) ( void * , uint32_t ) ;
void ( * re cei v e ) ( void * , uint32_t ) ;
void ( * close ) ( void ) ;
void ( * change_channel ) ( uint8_t ) ;
} hal_comm_dev_t ;
s t a t i c hal_comm_dev_t dev_radio = {
&radio_open ,
&radio_configure ,
&radio_send ,
&radio_receive ,
&radio_close ,
&radio_change_channel
};
s t a t i c hal_comm_dev_t dev_uart = {
NULL,
&uart_configure ,
&uart_send ,
&uart_receive ,
NULL,
NULL
};
Listing 4.2: HAL structure of communication drivers
The listing 4.2, communication structure of HAL library can be seen. The struct type called
hal_comm_dev_t is defined for all communication modules inside of the SOC. After investiga-
GPIO Device Busy-wait Device Timer Device Radio Device UART Device
Translated Device
Functions
Hardware Abstraction Layer
Native Driver
Functions
GPIO library Busy-wait library Timer library Radio Library UART Library
Board Specific/Abstraction Library
Hardware
Figure 4.2: Interaction between HAL and Hard Drivers and Structural Design
tion about communication protocols, model about most of the communication type has been
decided. According to this model, every communication module either have open, close func-
tionality such as radio module or just uses send and receive functions such as SPI, I2C and UART.
Also most of the communication chips allows or needs user to configure and some of them they
do not allow or require this. Therefore, as it can be seen the example below, if one of the func-
tions is unrelated or unimplemented for that communication protocols, this can be easily dis-
carded giving NULL value to particular field of struct. On the other hand, new hard drivers of
sensors or module must be designed and implemented according to HAL. Once it is developed
accordingly to that, it can be easily implemented and used inside of the operating system. The
other part of HAL are implemented as same way as communication and GPIO structs. Interac-
tion between HAL and hard drivers and structural design of them can be seen at below.
Chapter 5
Kernel Architecture
In this chapter, designing steps of kernel architecture, theory lies behind it and how kernel op-
erates inside of the particular operating system are explained in detail. Kernel has two main task
in this system. To control context switching and thus, achieve multitasking and schedule given
task according to their priority.
Kernel is the part of operating systems, which contains the central core of operating system
functionality. It has complete control over everything that occurs in the system. Usually it is
the first program loaded on start-up, and then manages the remainder of the start-up, as well
as input/output requests from software, translating them into data processing instructions for
the central processing unit. It is also responsible for managing memory, and for managing and
communicating with computing peripherals.
Target platform (cortex-M0) has a number of features targeting at embedded operating sys-
tem support. These are:
• System Tick Timer: It has a special timer called System Tick (SysTick) timer. It is 24-bit
counter that can be used to create needed system tick for some operation in kernel espe-
cially to control context switching. Since this is an optional feature, Nordic semiconduc-
tor, which developed our target platform, decided not to implement this timer. Therefore,
a real time counter is used instead of this special counter.
• Two stack pointers: There are two stack pointer in the processor architecture called The
Main Stack Pointer (MSP) and the Process Stack Pointer (PSP). With the help of having two
55
CHAPTER 5. KERNEL ARCHITECTURE 56
Figure 5.1: Multitasking and Context Switching[37]
stack pointers, OS kernel and application stack can be separated [37].
• A SVCall exception and SVC instruction: This exception or call can be used by applications
to access OS services and make development of OS easier [37].
• A PendSV exception: PendSV is an interrupt-driven request for system-level service. In an

OS environment, use PendSV for context switching when no other exception is active.
5.1 5.1. Multitasking and Context Switching
Operating systems that are used in embedded applications are design to handle with multiple
tasks. They achieve this dividing the processor execution time into numbers of tasks and per-
form each task in different slot. The length of these time slots differs from one operating system
to another and depends on hardware as well. Scheduler of the operating system works at the
end of each time slot to change the task or to decide changing of the task. That switching of
tasks is called context switching [37]. During the context switching, execution context (state pf a
process, task or thread) is stored and restored to resume the execution of switched process later.
Therefore, multiple tasks can share a single CPU and this is the essential part of multitasking op-
erating systems. Performing context switch and multitasking in a single processor system can
be seen at the figure 5.1.
To achieve context switching in periodic way, the task execution should be interrupted by a
hardware like a timer. When the timer asserts an interrupt, an interrupt handler or exception
handler must deal with task scheduling. The handler might also carry out other OS maintenance
Figure 5.2: Context Switching from one task to another[37]
tasks. In this particular operating system, systick library is developed for that purpose. User can
configure the interval of system tick timer from configuration file and default value is 400 Hz.
With default value, operating system can change task for four hundred times per second. In
order to keep overhead as low as possible, execution time of this handler must be very small.
Therefore, this handler only checks need for context switching and set some registers to main-
tain the whole system. To achieve context switching two stack pointers provided by Cortex-M0
architecture were used. One of them called MSP that can be used at start-up and in exception
handlers, including OS operations. Other one is known as PSP that is typically used by applica-
tion tasks in a multitasking system Both of them are 32-bit registers and but only one of them
can be used at once according to the vale in CONTROL special register and current mode. The
MSP is the default stack pointer and initialized after the reset by value from the first word of the
memory. Only MSP can be used in simple applications. However, multiple stack regions are
needed for embedded operating systems and high reliability. With the help of separation, one
stack could be used for OS kernel and exceptions and the other one for tasks. In addition, this
separation also leads to reduce the chance of stack error. Also using memory protection unit
can limit stack usage of task. However, in this project to keep simplicity and follow lightweight
design, it is not implemented. The kernel has to keep track of the stack pointer values for each
task during context switching and switch back to the PSP value to provide stack for each task.
During context switching, the SP for the exiting application task in the PSP will have to be saved
and the PSP will then change to the SP location for the next task. This process is illustrated at
the figure 5.2.
v o l a t i l e stack_type_t * p_top_of_stack ;
base_type_t priority ;
stack_type_t * p_stack ;
char task_name [ 3 ] ;
task_state_t state ;
uint8_t blocked_by_mutex_id ;
list_node_t p_node ;
} task_control_t ;
Listing 5.1: Task Control Structure
In this operating system, kernel has native libraries for task creation. Task library is respon-
sible from main task functionalities. In addition to that, stack function library provides memory
allocation and initialization functions for each task. The task control structure can be seen at
the listing 5.1. When a task is being created, task create function is called. Firstly it initialises
the lists for task states. Afterwards it calls one of stack functions called allocate stack and control
block to allocate required memory space and stack for particular task if there is enough space. If
there is not enough space for the task, it returns an error code and creation of task is cancelled.
After successfully allocate the memory for the stack and control block, other fields of control
structure has been filled with relevant values. Then if there is no running task currently, current
task pointer value assigned as particular task pointer. In order to reduce to size of the code, for
allocating memory to control pointer, our own simple memory allocation function was written.
After implementation of tasks, dispatcher function for context switching is implemented.
Dispatcher function is implemented into PendSV handler. It is asserted by setting a pending
status bit in the system control block in cortex-M0. Thus, it can be set even during the execution
of higher priority exception handler. PendSV exception is used to achieve context switching as
it is mentioned before. In addition, this exception can be used for the following:
• Separating an interrupt processing task into two halves.

Figure 5.3: Context Switch without PendSV[37]
• The first half must be executed quickly and is handled by a high-priority interrupt service
routine (ISR)
• The second half is less-timing critical and can be handled by a deferred PendSV handler
with a lower priority. Therefore, it allows other high-priority interrupt requests to be pro-
cessed quickly.
The SysTick exception is set up as a medium priority exception. Therefore, the SysTich han-
dler can be called even during the execution of another interrupt handler. However, the context
switching should not be discarded while an interrupt service routine is running. In that case,
the interrupt service could be broken into multiple parts. Thus, it leads to increase of overhead
and this can be clearly seen at the figure 5.3. In this particular system, systick check the need of
context switch and if context switch must be called, it sets PendSV exception and processor im-
mediately perform context switching since it has higher priority than system tick timer handler.
There are three native functions in dispatcher library. Dispatcher first kick function provides
the necessary first kick for the scheduler. It obtains the location of the current task control block.
Then saves the first item in control block as top of the stack. Afterwards it discards everything
up to register 0 (R0) and makes PSP as the new top of the stack to use in the task. Moreover, it
manipulates PSP stack and copy the contents. Finally, it pops out the program counter to jump
to the user defined task code.
Dispatcher switch handler, as it can be understood from its name, handles with contents
during the context switching. PendSV handler calls that function to switch between tasks. It
works same as first kick function. Only difference is saving and restoring registers for switched
tasks.
Last function of dispatcher is task yield function, which asserts PendSV exception. It is used
at the end of each task to indicate the end of task. With the help of yield function, scheduler
can choose the new task to execute. Flow chart of context switching and functional algorithm of
dispatcher and system tick timer can be seen at the figure below.
To complete the kernel of particular operating system, software interrupt mechanism is
needed to allow tasks to trigger specific kernel functions. That can be achieved using Super-
visor Call (SVCall) in Cortex-M0 architecture. This exception can be triggered using provided
instruction for SVC. . “Typically, when the SVC instruction is executed, the SVCall exception is
triggered and the processor will execute the SVCall exception handler immediately, unless an
exception with a higher or same priority arrived at the same time and is being served first” [37].
In some systems, SVC is used as access point of some operating system functions which do not
require any address information. Thus, it provides complete separation between application
and operating system. In this system, SVC is used as a gateway for using mutex and IPC.
Inter-process communication or commonly known as IPC is a mechanism or services, pro-
vided by operating service, which allows processes to share data with each other. In this par-
ticular project, IPC capability is limited due to light-weight architecture. Therefore, to achieve
this mechanism only mutex concept is implemented. Mutex is used abbreviation of mutual ex-
clusion object. In computer programming a mutex object allows tasks and process to share the
same resource not at the same time. When a program is started, a mutex is created with a unique
name. After this stage, any task/process that needs the resource must lock the mutex from other
task/process while it is using the resource. The mutex is set to unlock when the data is no longer
needed or the routine is finished. In this system, native mutex function is implemented inside
of the kernel since also other kernel functions may need to use mutex mechanism. User needs
to declare the mutex object as a structure to use it. Mutex structure can be seen at the figure 5.2.
System Tick Timer System

maintenance
Does Tick
Increase system
Counter equals Yes
time second
400?
No
Increase system
time second
Native Function called Check context

from Task Library switching
PendSv Exception
Context Switch Set PendSV Call Dispathcer

Yes
is needed? exception switch Handler
Store switched task

No and restore stack
for new task
Continue the
Start execution of
execution of current
new task
task
Figure 5.4: Flow diagram of system tick handler and dispatcher

mutex_state_t state ;
mutex_id_t mutex_id ;
bool initialized ;
} mutex_struct_t ;
Listing 5.2: Mutex Structure
The mutex object has very simple design. There are two possible states for mutex object and
these are locked and unlocked states. Other fields of mutex structure (mutex_id and initialized)
are set by operating system during the creation of mutex element. In order the lock or unlock
the mutex object, two functions with the same name are implemented. User just needs to give
specific mutex as the argument of the function to use them. In order the lock and unlock mutex
even from top level without having issue with other kernel operation, SVC exception is used.
Once the function is called to lock/unlock the mutex, this immediately asserts an SVC exception.
In order pass arguments to the this exception, master stack pointer(MSP) and process stack
pointer(PSP) are used as it was mentioned earlier in this section.
5.2 Priority-Driven Scheduler and Real-Time Compatibility
Scheduling is an important concept in operating systems. It means assigning of a work/task or

processes to a resource, which completes that. The work may be computation elements such as
task, process or thread which are in turn scheduled onto specific hardware such as processor,
peripheral unit or server. Schedulers allows user to use and share system resource effectively
and efficiently. Scheduling is fundamental to computation itself, and an intrinsic part of the
execution model of a computer or an embedded system; the concept of scheduling makes it
possible to have computer multitasking with a single processor systems.
In different systems and application, schedulers may have different goals. For instance, max-
imizing throughput (the total amount of work completed per time unit), minimizing response
time (time from work becoming enabled until the first point it begins execution on resources), or
minimizing latency (the time between work becoming enabled and its subsequent completion).
During the implementation of scheduler, usually these goals conflict with each other. Therefore
suitable compromise for different application must be done. Preference is given to any one of
the concerns mentioned above, depending upon the user’s needs and objectives. In real time
application like embedded systems for control industry, the scheduler must ensure that all the
task or process can meet deadlines. Scheduled tasks can also be distributed to remote devices
across a network and managed through an administrative back end. Since this particular project
is based on lightweight design for embedded systems, the main goals of the scheduler are simple
and real-time compatible task execution.
Priority levels can be defined for each task to execute high-priority task before lower priority
tasks. In addition, the operating system run the task for a number of time slots until it reaches an
idle state, if the task has a higher priority than others do. On the other hand, exception priority
and priority definition in the operating system is completely different than each other. The
definition of task priority is based on application, user and sometime Operating System design
and varies between different application and operating systems [37].
Fixed-priority pre-emptive scheduling is the base of particular scheduler. In this scheduling
schemes, the scheduler ensures that the highest priority task is executed among the all tasks
that are currently ready to execute in scheduler. Since it is also pre-emptive scheduling, system
tick provides the scheduler to switch after the task has had a given period to execute. With
that scheduler ensure that no task is executed in processor for any time longer than system
tick. Nevertheless, this leads system to lockout since priority is given to higher-priority tasks,
the lower-priority tasks could wait an indefinite amount of time. Therefore, user must know
the application and assign priorities carefully. Most of the embedded operating system use pre-
emptive schedulers.
In this particular project, scheduler can be considered as the conductor of an orchestra.
Scheduler does not provide lots of functionalities to user or the other layers of operating sys-
tem. However, it orchestrates the system and task. Thus, this system has very small and efficient
scheduler using pre-emptive scheduling. Most of the priority and task management functional-
ities are provided by task library.
Every task has a state to be executed properly by scheduler. This system a task may have six
different state to ensure reliable execution of task. These are running, ready, waiting, sleep, sus-
pended and blocked. Task library also provides lists about task states. Whenever a task changes
Task Create Function
Creation of
the Task Get highest Ready
priorty functon
Mutex API unblock function
Task API
Wake Up Function
Switch context or
get highest priorty function
Waiting Blocked
Creation of higher priority task

Mutex API task block function
Running
Task Yield Function
Task API
Sleep Function
Sleep Suspended
Figure 5.5: Transition of Task States
its state, it is removed from previous state’s list and added to current state’s list. Transition be-
tween states can be seen at figure 5.5. When task is created, its state is set to ready state. In the
same creation function, priority control is done and if it is the only task to be scheduled or it has
the higher priority than current task, its state is changed immediately to running state.
During the context switching relevant functionality is also provided by task library. If context
switching occurs, current task’s state is changed to waiting and it is added to waiting list. After
the execution of highest priority task, task yield function must be used. With yield function,
current state is declared as suspended state and highest priority task can be chosen. Whenever
a new task must be chosen after the execution of highest priority task, firstly ready list is checked
System Tick Occurs
Systick
Handler
Is there any task which has
higher priority than current
Return to current task
task in ready list?
Assert PendSV
interrupt
A task
Remove current event
from ready list and
add it to waiting list Execution of task
Is it the end of
Change the state of execution?(Task
swithced task yield)
Dispatcher
Swtich
handler
Store high and low Suspend task

register for swithced
task
Find the highest

priorty function and
Switch context wait for system tick
Change state and the

list of higher priority
task
Scheduler
Figure 5.6: The flow chart of priority-driven scheduler during the context switching
for same priority. If there is no task in ready list same priority with the highest priority task, then
waiting list could be started to check with the same priority. Whenever a suitable task is found to
execute, its states changed to running again and all process repeats after the execution of every
task. The flow chart of priority-driven scheduler during the context switching can be seen at the
figure 5.6.
In order to create more reliable semaphores and prevention of deadlock, block state and
list is added to task library. Task block function is called by mutex function, if particular mutex
which is tried to locked by task is already lock. Whenever the mutex object is unlocked, the
functions unlock the task, which was blocked by mutex. This block and unblock functions is
called by SVC as it was earlier mentioned in this paper.

Moreover, in order to sleep and wake the task according to an external event or interrupt,
sleep task and wake up task functions are implemented. Task is put to sleep state whenever
sleep function is called and can be woken up using wake up function and add immediately to
ready list. However, when it wakes up, if there is a higher priority task that is being executed by
scheduler, the task waits the end of higher priority task’s execution.
Chapter 6
Other System Components
In this particular section, other small components of operating systems are explained. As it was
mentioned before, operating systems are formed by many components. Although hard drivers,
hardware abstraction layer and kernel form the main part of the operating system, there are
some components, which are implemented in order to develop software application using this
particular operating system. These components are utilities library, application-programming
interface and user defined function (application layer).
6.1 Utilities Library
In order to make development easier, particular operating system provides some basic function-
alities for basic software concepts. The provided libraries are linked list and memory allocation
libraries. In computer science, a linked list is a linear collection of data elements, called nodes
pointing to the next node by means of a pointer. This structure formed by a group of nodes
that together represent a sequence. Each node contains data and a reference to next node and
it provides efficient insertion and removal of elements from any position in an easy way. The
structure of basic linked list is illustrated at the figure 6.1
This structure is the one of the simplest and most common structure in computer science. It
can be used to implement abstract data types, including list, stacks, queues, associative arrays,
and S-expressions.
Linked lists are among the simplest and most common data structures. They can be used
67
CHAPTER 6. OTHER SYSTEM COMPONENTS 68
Figure 6.1: Structure of Linked List [38]
to implement several other common abstract data types, including lists (the abstract data type),
stacks, etc. The purpose of implementing a linked list rather than an array is the insertion and
removal of the element can be done easily without needing reallocation or reorganization of the
entire structure since the data items does not need to be stored in the memory. Linked lists allow
insertion and removal of nodes at any point in the list, and can do so with a constant number of
operations if the link before the link being added or removed is maintained during list traversal.
On the other hand, simple linked lists by themselves do not allow random access to the data,
or any form of efficient indexing. Thus, many basic operations such as obtaining the last node
of the list (assuming that the last node is not maintained as separate node reference in the list
structure), or finding a node that contains a given datum, may require sequential scanning of
most or all of the list elements.
Linked list library provides four function to user to create reliable linked lists. Every list must
be initialized before adding new nodes and list_init function is used for that purpose. All
the content in the particular list can be deleted by using list_clear function. In order to add
a node to initialized list, a node must be declared using node data structure in the library. Af-
terwards particular node can be added or removed bylist_node_add and list node_remove
functions. In some kernel functions in this system, linked list library has been used and it can
be used also in application layer for user specific functions.
Memory management is the concept of managing system memory in kernel or system level.
Essentially providing functions to allocate portions or sections of the memory for the programs
at their request and free that section after the usage is required.
There are several methods and concepts which is implemented in computers. Virtual mem-
ory systems separate the memory addresses used by a process from actual physical addresses,
allowing separation of processes and increasing the effectively available amount of RAM using
paging or swapping to secondary storage. However in this particular project none of them are
used since using this concept would be overkill for embedded oriented operating system. There-
fore, basic memory allocation function is implemented for developers to use it in the application
which memory allocation is needed.
Allocation request contains locating a block of unused memory of sufficient size .These re-
quest are done by allocating section from the heap. At any given time, some parts of the heap
are in use, while some are "free" (unused) and thus available for future allocations.
On the other hand, there are several issues which makes implementation complicated. For
instance external fragmentation may occur when there are many small gaps between allocated
memory blocks. Allocations must be tracked to ensure that they do not overlap and no memory
is lost as memory leak.
Since standard memory allocation library has large footprint, a simple memory allocation
utility library was developed. This library has basic functionality for memory allocation. Mem-
ory allocation can be achieved using malloc_alligned function. It allocates memory and re-
turns the pointer of the stack. In addition, allocated memory can be freed using mfree function.
These functions are used in creation of the task for the operating system in kernel level and can
be used in user applications as well. At the figure 6.2, memory sections can be seen after the
allocation.
6.2 Application Programming Interface
Application Programming Interface commonly known as API is concept, which contains set of
routines, protocols and tools to develop software applications. Therefore, APIs can be seen
translator for a software or hardware component in terms of their implementations that allows
definitions and implementations to vary without compromising the interface. Therefore, a good
Figure 6.2: Main memory after the memory allocation [39]
API provides all the building blocks of particular component and makes software development
easier.
An API can be developed for the component of operating system, web-based system or
database system, and provides functionalities to user to create application for that particular
system by using specific programming language. Depends on the programming language, usu-
ally an API comes in the form of a library, which contains specifications for routines, data struc-
tures, object classes, and variables. In the operating systems, API allows user/developers to use
predefined functions to use or interact with the operating system component instead of writing
them from scratch.
Almost all the operating system in the market such as Windows, Unix and the Mac OS pro-
vide an API for developers. Moreover, it can be used by game consoles, other hardware specific
devices or embedded systems which can execute software programs. For instance, in order to
develop application for Android can use an Android API to use the front camera of an Android-
based device. Although developer can develop library for camera from scratch using native An-
droid libraries, it would take lots of effort and would not be compatible with every android de-
vice.
This particular operating system provides five different APIs to user for development usage.
These are boot, communication, IOT service, IPC and task APIs. In order to keep simplicity and
achieve lightweight design, only the most needed APIs have been implemented to ease the pain
Application Layer
API functions
API
API IOT Service API Task API IPC API Boot
Comm
Native Kernel
Functions
Comm Radio Stack Supervisor

Task Mutex
Devs Dev Functions Call
Kernel
Figure 6.3: The structures and functionality of API
of development process. For other functionalities of hardware, user can directly interact with
hardware abstraction layer. Since it is an embedded operating system, further development of
APIs is not required. The structures and functionality of API can be seen at the figure 6.3.
Booting or commonly known as booting up is the initialization routines of the electronics
and computer systems. This process can be executed after CPU is switched from off the on
(hard) to detect hardware error or can be executed by a software command (soft). When boot-
ing is complete, the normal, operative and runtime functionality of operating system is attained.
The piece of software, which is responsible from booting-up, is called boot-loader. It loads the
operating system or some system software. On modern personal computers, this process may
take tens of seconds and usually helps to perform a self error test, locate and initialize periph-
eral devices. However in the embedded systems, the booting sequence is not that complicated.
Simply, it sets hardware registers and run operational programs that are stored in ROM. There-
fore, also in that particular operating system booting sequence is short and small. API boot
provides boot start function to user. User needs to call that function to set the operating system
and the platform correctly. Boot start function creates a task and executes it before the user de-
fined tasks. It sets GPIO, communication block and other small components in the system and
it returns error message if there is a hardware error in the platform.

The other API library provided by OS is API communication. Actually, this is the only API
which is not provided for user. It is provided for IOT service API to make the development of
that component easier and compatible with further release of the operating system. User can
use other communication protocol or modules such as SPI,I2C and etc. by using hardware ab-
straction layer. However the privilege to use protocols or modules which can allow device to
communicate with the other devices, is not given to user. User can interact with IOT service
to have basic functionality in this issue. IOT service and its API are explained in detail in the
section called “IOT Service and Security”.
Since mutex element is created by using the native functions and supervisor call of the ker-
nel, API IPC is implemented to secure the kernel from user-defined task errors or mistakes. API
task provides basic functionality of the mutex object. User must need to declare a mutex object
using the given data structure type in the api_ipc.h. After the declaration of the object, user
can lock and unlock the mutex element by using basic functions with the same name. API mu-
tex functions are executed in the same way with mutex native kernel functions. Only difference
is using API, the important details are hidden from users/developers and basic functionality is
provided.
The most important and the biggest API in the particular system is API task. As it was dis-
cussed before, the most important feature for embedded operating systems is task execution.
The task execution needs to have real time compatibility and should be performed reliably and
efficiently. Therefore, the API for task creation and scheduling must be user-friendly. In addition
to that, it needs to give the full functionality to user with a small size design. The native library
of the task was explained earlier section in this paper. The API task gives user the every aspect
of task scheduling and creation. It provides functions for creation and deletion of tasks. Before
the creation of the task, initialization function must be called to set the operating system for this
API. User needs to use yield function if scheduler wants to be run after the execution of the task.
Moreover, user can start scheduling of the user tasks by using api_task_start function. These
functions may return error message, if unexpected error occurs during the task creation, dele-
tion or starting scheduler. Furthermore, this API provides sleep function depends on specific
event such as timer event in the operating system.
6.3 User defined functions
Application program means a piece of code or software, which is designed to perform a group of
functions, task or activities to achieve specific user task with using particular operating system.
For instance in personal computer worlds, a word processor, an accounting application or a web
browser are considered as applications. The concepts refers to all kind of applications running
in specific operating systems. Applications may be bundled with the computer and its system
software or published separately, and may be coded as proprietary, open-source or university
projects.
An application is designed to help user perform an activity. Therefore, it differs from an
operating system (which runs a computer or embedded system) or a utility (which performs
maintenance or general-purpose functionality). Depending on the activity and targeted operat-
ing system for which it was designed, an application can manipulate text, numbers, hardware,
or a combination of these elements.
The delineation between system software such as operating systems and application soft-
ware is not exact, however, and is occasionally the object of controversy [40] .For instance;
there are debates about whether Microsoft’s Internet Explorer web browser was part of its Win-
dows operating system or a separable piece of application software. As another example, the
GNU/Linux naming controversy is, in part, due to disagreement about the relationship between
the Linux kernel and the operating systems built over this kernel. In the embedded operating
systems, the application can be indistinguishable to the user, since user can be also developer
and they can develop software to control piece of hardware.
However in this particular operating system, is specifically designed for low-level applica-
tions. Therefore, every other piece of code which is written using operating system API is con-
sidered as application software. Moreover, user needs to do programming in C language since
there is no internal compiler or provided graphical user interface. Thus, the main function is left
for user applications. User needs to include specific API header files and call some functions to
run the platform. Examples about simple application layer can be found in appendix. Moreover,
in order to configure operating system specifically for application, a configuration file is imple-
mented. It allows enabling and disabling functionalities of operating system to user to provide
flexibility for different applications. The configuration file can be seen at the listing 6.1.
# define XOS_H
#include < s t d i n t . h>

#include < s t d l i b . h>
//memory configuration
# define configMINIMAL_STACK_SIZE ( ( unsigned short )600)
# define configTOTAL_HEAP_SIZE (( size_t )(3000))
// t a s k s configuration
# define TASK_PRIORITY_NUMBER 16
# define IDLE_TASK_PRIORITY 15
# define IOT_SERVICE_PRIORITY 1
// system t i c k configuration
# define RTC_FREQUENCY (400UL)
// API configuration
// api_task configuration
# define TASK_STACK_DEPTH 200
// a p i _ i o t _ s e r v i c e configuration
# define IOT_SERVICE (1)
# define IOT_SERVICE_MANAGER (1)
# define IOT_SERVICE_NODE (0)
# define IOT_SERVICE_SUPPORTED_NODE_NUMBER (255)
// i n t e r r u p t handlers
# define xPortSysTickHandler RTC0_IRQHandler
# define dispatcher_switch_handler PendSV_Handler
// hal configuration
# define BUSY_WAIT_DEV (1)
# define GPIO_DEV (1)
# define I2C_DEV (0)
# define SPI_DEV (0)

# define UART_DEV (1)
# define RADIO_DEV (1)
# define ADC_DEV (0)
# define TIMER_DEV (1)
# define COM_DEVS (1)
extern v o l a t i l e uint32_t system_time_second ;

# endif
Listing 6.1: Configuration File of Particular Operating System

Chapter 7
Radio Library, Stack and Protocol
For this particular project, in order to provide wireless communication, a new simple protocol
has been design. In addition to that, in order to have efficient communication while using that
protocol, radio library is designed from scratch. Library and Protocol design are explained in de-
tail in this section. Implementation and purpose of developing a new library, stack and protocol
is mentioned as well.
As it was explained before targeted platform(NRF51822) has a radio module in the same SOC
together with an arm Cortex-M0 microcontroller. Nordic Semiconductor provides a software in
order to control radio module by using Bluetooth Smart protocol. The piece of software is called
Soft-Device, which is precompiled and linked binary software implementing a Bluetooth 4.1 low
energy protocol stack for the nRF51 series of chips.
They also provide API which is written in C language and which provides the application
complete compiler and linker independence from implementation. This allows user to develop
their project or application as using a standard arm Cortex-M0 device without needing to inte-
grate with proprietary chip-vendor software frameworks [41]. Structure of Soft-Device can be
seen at the figure 7.1.
Besides the advantages of Soft-Device in Nordic products, this Soft-Device is not imple-
mented in this particular project. Firstly, Bluetooth stack has a very large footprint. Since the
main purpose of project to develop a small size embedded operating system, implementing the
large Bluetooth stack would be inappropriate. In addition, there is no need to have different
profiles for different applications as Bluetooth protocol provides. This particular project, pro-
76
CHAPTER 7. RADIO LIBRARY, STACK AND PROTOCOL 77
Figure 7.1: Structure of Soft-Device[41]
vides functionality and task execution for wireless sensor networks and internet of things de-
vices. User or developer can develop their own project easily on top of this particular operating
system. On the other hand, since this soft-Device comes with pre-compiled way and API, there
is hard to see what is going on inside of that pre-compiled software. They may be undesired
behavior for developers in some specific applications.
As it was detailed before, particular SOC contains very flexible and reliable 2.4 GHz Radio
transceiver. In order to develop specific library for this module features and details of radio
modules must be investigated. First of all, a DMA module which is called Easy-DMA, is imple-
mented by vendor into radio module. DMA module provides reading and writing data packets
from and to the RAM sections without interrupting CPU. It has PACKETPTR pointer for receiving
and transmitting packets. Processor needs to reconfigure this pointer each time before starting
of module via START task command. Accessing the RAM will have been finished by DMA mod-
ule when radio DISABLE task is executed. If the PACKETPTR is not pointing to the Data RAM
region, an Easy-DMA transfer will result in a Hard-Fault [35].
A Radio packet contains the following fields: PREAMBLE, ADDRESS, LENGTH, S0, S1, PAY-
LOAD and CRC as illustrated in Figure X. When radio module sends a packet, order of the packet
Figure 7.2: On-air packet layout [35]
is described from left to right in the figure 7.1. The pre-amble bits is sent firstly on-air. It is al-
ways one byte long for all the modes provided by MODE register.
Radio module also has CRC generator to calculate cyclic redundancy check over the whole
packet excluding preamble. Therefore, no need to implement an external CRC library in order
to check CRC errors. In addition to that, CRC error is detected by radio module automatically.
After setting packet and radio register, series of command must be followed to send the packet
to other device. The module must first ramp-up in TX mode. This mode can be set when TXEN
task command is triggered .After the successfully ramping up of radio module, it generates the
READY event to indicate that transmission can be started [35]. However, event is not used in
this library as an interrupt in order to keep reliability. System simply busy-waits while radio is
getting ready. Afterwards, transmission is started by using START task.
Like in transmitting, radio module must be ramped-up in RX mode to be able to receive a
packet. Once the RXEN task is executed, RX ramp-up sequence is started. After the successfully
ramping up of radio module, it generates the READY event to indicate that packet reception can
be started [35]. In this particular event called packet-received event, radio module asserts an in-
terrupt. System does not wait to get a new packet; it performs other task or applications. Once
it receives a packet, it asserts and interrupt and radio stack and library will handle with packet.
Packet handling is implemented on IOT service and more information about the service can be
found in the following chapter. After asserting packet receive interrupt, radio module enters RX-
IDLE state and radio module is closed/disabled by radio library to reduce energy consumption.
Particular radio modules supports communication between 2.400 GHz and 2.483 GHz since it
is specifically designed for Bluetooth Smart protocol. Therefore, new protocol and library use
same physical layer with Bluetooth protocol.
In this particular protocol, defines two types of device. These are manager and node de-
vices. User needs to assign the device one of them. Manager communicates every device in the
cluster and analyze their need and enable to share resource and information inside the cluster.
Every cluster can only have one manager. On the other hand, node device is the other device
which performs specific task such as computing, sensing or actuating. Nodes do not commu-
nicate with each other. According to protocol every device that is connected to cluster, uses
different channel to communicate with manager. A node is registered by manager at the be-
ginning. Manager has a choice to reject the node or accept to the cluster. This communication
between unregistered node and manager is done at the 2.44 GHz. After successfully registration
of the node, manager assigns specific id and channel for every particular node/device. Each
new channel is given with 1 MHz difference from previous assigned channel. Therefore, this
protocol only support 82 device inside of one cluster. Bandwidth for every channel is 1 Mb/s.
More information about manager and node devices functionality are explained in detail in next
section.
The radio library has simple and small architecture. It provides six different function to hard-
ware abstraction layer for basic functionalities of radio module. These are open, close, send,
receive, configure and change channel functions. Radio module is configured by the operat-
ing system at the start-up of the platform using boot function. Before sending and receiving
packet radio_open function must be called to start ramping up task for transmission or packet
reception. Radio_change_channel function is responsible from changing channel frequency
for particular device. User needs to give message pointer and length as argument before using
send and receive functions. Only receive function creates an external interrupt which is han-
dled in IOT service layer. Radio module has a register called state register to use radio state in
callback functions. However Nordi Semiconductor did not implement this register yet [34] . In
order to achieve state tracking in this system a state type is declared separately and a variable is
declared from it. Every function of radio module, this variable is changed according to datasheet
to track state of radio module and send it to the high-level callback function. The structure of
radio library and interaction between radio library and IOT service can be seen at the figure 7.3.
A communication protocol means a group of rules that provides two or more devices to
transmit and receive information to each other via any kind of variation of physical and elec-
tronics quantity in tele communication. The set of rules defines the syntax, semantics and syn-
IOT Communication
API Boot
Service Layer
Radio
Device
Callback Configuration
Radio Library
Radio Module
Figure 7.3: Structure of Radio Library and Interactions
chronization of communication. Protocols may be implemented by hardware, software, or a

combination of both [42].
Systems use defined formats for exchanging messages. Each message has an exact meaning
intended to elicit a response from a range of possible responses pre-determined for that partic-
ular situation. Implementation is independent from specified behavior of protocol. In addition,
there is a close analogy between protocols and programming languages since protocols are to
communications as programming languages are to computations [43].
In computer and embedded systems, this protocols is expressed by algorithms or different
data structures. Expressing the algorithms in a portable programming language makes the pro-
tocol software operating system independent. Typically, a set of processes, which manipulates
shared data to communicate with each other, is provided by operating systems. This commu-
nication is achieved by using well implement protocols which may be implemented into the
process code itself. On the other hand, since there is no shared memory among devices, devices
needs to communicate with each other using a shared transmission medium. Each device or
system may use different hardware or operating system while using same protocol.
From Node to Manager
Node Registired Node Message Message( Information Mail/

ID(factory ID) Node ID Message Type Length(n) Publishing Ad)
1 byte 1 byte 1 byte 1 byte n byte
From Manager to Node
Registired Manager Message

Manager ID
Node ID Message Type Length(n)
Message
1 byte 1 byte 1 byte 1 byte n byte
Figure 7.4: Packet Layout of the New Radio Protocol
This particular protocol is developed in order to enable resource and information while
keeping protocol stack footprint as small as possible. This protocol basically is based on two
different concept and these are yellow pages and mailing system. Every node creates a mail
periodically and interval of that period can be configured by user/developer. Manager collects
these mail whenever it Is suitable for it and create yellow pages according to need of the nodes.
Moreover manager can calculate the workload of the cluster or particular node. All relevant in-
formation to analyze particular node is included to the mail with timestamp. To achieve this
conceptual communication, the protocol is designed based on this. Every node has one byte
node id which is given by manager and that is the first byte of the message from node to man-
ager. Afterwards one byte device specific id (factory id) is included. Then one byte message type
and one byte of message length comes. Message Length * 1byte long message comes afterwards.
The protocol from manager to node is similar as well. Only first byte of message contains man-
ager id and second byte of message contains node id which is given by manager. The packet
layout of particular protocol can be seen at the figure 7.4.
A protocol stack means an implementation of a computer networking protocol suite. The
communication protocols can be considered as definition and protocol stack can be considered
the implementation of particular rules in software.
Protocol stacks are often designed as layered software. This modularization allows user to
User
API IOT 7)Application Layer
application
6)Presentation Layer
5)Session Layer
IOT Service
4)Transport Layer
3)Network Layer

2)Data Link Layer
Radio Library
1)Physical Layer
Radio Module(hardware)
Figure 7.5: The stack structure of the radio protocol based on OSI model
design and evaluation easier. The lowest layer of protocols usually handles with low-level, phys-
ical interaction of hardware. In this case (this particular operating system), lowest layer of pro-
tocol stack is radio library. Higher layers adds more features and most of the user applications
only deal with top layers.
There are three major section usually in every protocols stack. These are media, transport
and application layer. In addition to that, mostly two interfaces are implemented between those
layers by a particular operating system or platform.
This particular protocol is implemented using three different components of the system.
If seven layer OSI model considered as the base of this protocol, physical layer and data link
layer is performed by radio library. IOT Service handles with network, transport, session and
presentation layers. User can use the stack using IOT API and that forms application layer of the
protocol. The stack structure of the radio protocol based on OSI model is illustrated in the figure
7.5.
As it was discussed earlier in this section, main reason of the developing a specific radio li-
brary and protocol is large footprint of provided Bluetooth soft-device. This pre-compiled soft-
device needs 92 kb space in the flash memory. Moreover when it is enabled, it needs minimum 6
kb and regularly it requires around 8 kb space in the RAM. Furthermore, its maximum call stack
usage at run time is 1536 bytes. On the other hand, after successfully implementation of particu-
lar radio library and protocol, all the implementation takes 1,86 Kb of RAM and 2,67 Kb of Flash
memory. This also includes the size of IOT service, which enables resource and information
sharing. Since a running task handles with radio module and protocol, its maximum call stack
usage at run time is 600 bytes. Performance of the protocol and implementation and developing
radio library and protocol from scratch is discussed in the section called “Discussion”.
Chapter 8
IOT Service and Security
In this chapter, designing steps of internet of things service, theory lies behind it and how this
service operates inside of the particular operating system are explained in detail. Since IOT ser-
vice enables wireless communication and resource and information sharing among the cluster,
security vulnerability caused by this and how the security issue has been solved using this par-
ticular service have been mentioned.
In last twenty years, many technologic achievements have been happened in the Integrated
Circuit industry. These have led to develop small, energy efficient and faster processors and mi-
crocontrollers. Nowadays even a small microcontrollers which cost ten to twenty dollars, have
enough CPU power and memory space to achieve complicated tasks such as managing with
TCP/IP stack or Bluetooth communication. Even some operating systems can be supported by
this kind of low cost microcontrollers. Because of that lots of availability of low cost devices,
every kind of embedded systems have started to connect with each other and internet.
This technological achievements lead industry to use low cost embedded devices to create
large sensor network or IOT devices. Therefore even smallest failure or attack from outside have
turned out to be the most crucial problem since these devices either store private data or doing
very crucial task in harsh environment. Failures may not create big problems. However some-
times these kind of failures may cause of shutting down whole network. In order to prevent
network such a scenarios, a new approach just for sensor networks must be developed in both
network layer and software layer.
As it was discussed earlier in this paper, a new operating system service called Internet of
84
CHAPTER 8. IOT SERVICE AND SECURITY 85
Things Service has been introduced in this project to solve particular problems that are men-
tioned at the paragraph above. This service sits top of everything in this particular architecture
ad enables resource sharing and device connectivity. Application layer (user defined task and
functions) cannot reach communication layer without using IOT service. This service or archi-
tecture can be investigated from three different point of view and these are network and protocol
details, IOT service functionality and security issue.
8.1 Network Structure
Many different approaches have been tried and implemented so far in academy and industry
for sensor networks. However almost all of them had tried to improve the power consumption
or tried to allocate sensor automatically. On the other hand, our aim is creating solid sensor net-
work which can detect network faults, fix them and in the meantime it should keep the latency
as low as possible [44].
To implement and develop such a network, computer clustering can be the base of the in-
frastructure. Computer cluster consists of a set of loosely or tightly connected computers that
work together so that, in many respects, they can be viewed as a single system. Unlike grid com-
puters, computer clusters have each node set to perform the same task, controlled and sched-
uled by software. The computer clustering approach usually connects a number of readily avail-
able computing nodes via a fast local area network. The activities of the computing nodes are
orchestrated by "clustering middleware", a software layer that sits atop the nodes and allows the
users to treat the cluster as largely one cohesive computing unit [44].
In this particular project, different type of architectures for sensor networks and IOT devices
are investigated and suggested. Three different approach has been suggested and investigated
in detail. For all these three approaches, computer clustering can be considered as the base of
them. These structures are;
1. Classical Cluster Approach:
In classical approach for sensor network, sensor nodes perform their task coordinately
and send their data to main computer or to another node of the cluster which is respon-
sible for communication with main computer. Classical cluster approach can be seen at
Main Computer
Cluster1 Cluster2
Cluster Cluster
Manager Manager
c
c
Figure 8.1: Classical Cluster Approach
figure 8.1.
2. Cluster with Main Cluster Approach:
Classical cluster approach has some problems. Especially it is vulnerable some network
faults. These problems will discuss later on. Therefore, this approach needs to be im-
proved. Therefore, a node that should act as a kind of manager can be added. This man-
ager can analyze network health, rule the other nodes and send all data of cluster to main
computer. In addition to that, a manager cluster can be added to network which works like
manager node as well. Nodes may be connected to each other consecutively. This would
create communication fault. With these manager nodes, this problem will be solved. This
structure can be seen at figure 8.2 [45].
3. Cluster with Nodes Net Approach:
Different concept can be implemeted (collective intelligence, self-awareness, etc.) in or-

der to deal with every kind of network problem for wide variety applications. These con-
Cluster1
Cluster
Manage
r
Cluster1 Cluster2 Cluster2

Cluster Cluster Cluster
Manager Manager Manager
c c c
Figure 8.2: Cluster with Main Cluster
cepts requires different network structure. Also for hardware and software parts will be
challenging. However, hardware and software parts have not discussed yet. In order to
implement that kind of system, information about network and the other nodes must be
shared with each other. With shared information, nodes will get information about envi-
ronment and network. In addition, they can analyse the network and predict future errors.
Moreover, they can change their tasks or priority. That kind of system will act more like an
intelligent network. Structure of this approach can be seen at figure 8.3 [20].
Pros and cons of suggested structures have been investigated. In addition, possible prob-
lems and solutions of them are mentioned at the tables 8.1 and 8.2.
After investigation of cluster approaches, it can be easily seen that “classical approach” is
vulnerable to some problems. On the other hand, the approach called “Cluster with Nodes Net”
may be seen as the best approach among three of them. However, in the implementation part,
it would take more time for development and require large footprint in the software. Moreover,
nodes communicate with each other and manager in this approach. That would lead to increase
Cluster1
Cluster
Manage
r
c
Main Computer
Cluster1 Cluster2 Cluster2

Cluster Cluster Cluster
Manager Manager Manager
c c c
Figure 8.3: Cluster with Nodes Net
power consumption in each device. Since particular embedded operating system is focused
on low power consumption as well, implementation of this approach would be inappropriate.
Therefore, “Cluster with a Manager Approach” was decided to implement. The disadvantages
caused by this approach is tried to be solved by developing a reliable IOT service task and man-
ager device.
Implementation of this approach was divided in to two part. These are protocol which pro-
vides reliable and compatible communication between devices and IOT service which provides
efficient execution of wireless communication and resource and information sharing. There are
two types of device type in this particular approach and this also is implemented into protocol
and device configuration. These are called manager and node device types. Each device has
different execution sequence and algorithms according to its device configuration and as it was
explained before they have slightly different message layout in the protocol.
Approaches
Classical With Main Main Cluster + Node Net
Approach Cluster Approach Approach
-Simple and Solid -Relatively simple -Shared data can lead to
-Can be easily to implement collective intelligent
implemented to large scale -Immune for some and self-awareness.
Pros -Doesn’t require too network problems -Reconfigurable
much computation -Secure for outside -Immune for network
or process power network attacks problems
-Secure for outside -May decrease the
network attack latency
-Hard to implement
-Breaking down of the -Overhead may occur
-Not reconfigurable (self healing) manager node or -Needs more processor
Cons -Simple but not optimized cluster create big computing power
-Vulnerable to network problems big system problem. -May need different
-Not an intelligent system -Overhead may occur. hardware for cluster
-Not reconfigurable managers or special
nodes(for nowadays)
Table 8.1: Pros and Cons of Suggested Strcutres.
8.2 IOT Service
During the background research for this particular paper, problems and missing concept in the
embedded operating systems for sensor networks and IOT have been investigated and listed at
the earlier in this paper. Most of the embedded device use the old TinyOS or FreeRTOS. However,
none of them is developed specific for IOT or sensor networks. They may have lots of supported
architecture and they are good at real time applications. However, classical way of thinking
and developing OS and firmware caused those problems. As Einstein said «we cannot solve our
problems with same thinking we used when we created the». A new kernel architecture and
secure API must be achieved.
This particular service uses different parts of components and different module form hard-
ware. It is the biggest layer of the operating system after kernel. As it was discussed before it
allows device to communicate with the other devices. In order to achieve this purpose, it uses
radio module which is provided by communication layer. In addition to that, it has right to ma-
nipulate input and output of the device by using GPIO device provided by hardware abstraction
Problems Solutions
-Deployment of clusters will be
-It must be compatible for automated.Also application aware
every application. middleware may be solution for this
problem
-Vice/Second manager can be assigned. On
the other hand, every node in side of that
-Breaking down of manager node might
cluster can get manager different priority at
cause fatal error in the system.
the beginning. If manager breaks down,
according to their priority they can their
hierarchical position.
-Health report can be sent to main pc
-All nodes can start to breakdown at same
periodically. If connection goes down, node
time.
can be restarted
-Latency may increase if some nodes go -Cluster can be reconfigurable while
down. network is working
Table 8.2: Possible Problems and Solutions of Suggested Structures
layer. Moreover, since it is an operating system service, it is created by kernel as a task. Although
it is executed as regular task by kernel, since kernel contains a priority-driven scheduler, it has
highest priority. Therefore, it cannot be pre-empted by other task created by user. The software
structure of IOT service can be seen at the figure 8.4.
The particular approach called “Cluster with a Manager Approach” introduces two different
device types to network. Thus, this service comes with two different configuration. One for
manager devices and another one for node devices. Both of the device types have different
functionalities and execution procedure to follow. User must carefully configure IOT service by
using configuration file.
A cluster that is formed by using this protocol and operating system supports only one device
manager. Therefore, a device manager is unique for every cluster. Cluster manager communi-
cates with every device in the cluster. It is the only device in cluster, which can communicate
with the devices outside of the cluster or connect directly to Internet. It collects information
from every device and asses this information to calculate cluster and network workload. More-
over, it ensures that every node in cluster keeps its connectivity. It can also control specific node
using special function to control node resource such as computing power, memory or actuators
Information and resource

sharing functionality
IOT Service Layer IOT API
Radio
component
Task
Communication funtions
Layer GPIO
device
Kernel
Radio Device
Figure 8.4: IOT service software structure
like motors or LEDs. In addition, checking device connectivity and information collected from
nodes, a manager can check the problems in the cluster and do maintenance task for network.
Although its massive workload, since efficient execution and usage of task and components of
OS, it can also perform user specific tasks.
On the other hand, node devices act differently than manager devices. First of all, the IOT
service which runs in node devices, has less workload. Therefore, these devices can be used
for real-time application and they will provide better performance for meeting hard deadlines.
A cluster can have many nodes. However, the protocol sit top of this IOT service, support 82
nodes excluding manager device for each cluster. It creates information about itself periodically
and it sends to manager whenever manager ask to send information. It is able to communicate
only with manager. During the registration, node is assigned to a channel between 2.400 and
2.483 and only manager can change this assigned channel. Since it can only communicate with
manager, a node device consumes less power than manager device. Manager can use it as a
slave. However, this slavery usage of a node does not affect execution of other tasks. In addition,
IOT service of node devices has slightly smaller footprint than manager device’s IOT service,
since it has less functionality.
There are two concepts based on distributed systems and implemented into this structure.
These are yellow pages and mailing system. The name of the yellow pages is choose to de-
scribe of the structure in better way. Yellow pages can be considered as advertisement page
called yellow pages in newspapers. In this particular system, yellow pages are implemented into
only manager devices. Manager devices retrieves information from the nodes and creates these
pages. Manager can check yellow pages and decide which node needs what kind of resource.
Only needs of nodes are written in these pages. Therefore, one of the task of the manager is to
provide required source by nodes. Required source could be provided directly from manager
or it can be provided one of the node, which is not using particular resource currently. Yellow
pages are implemented as array of structures. Therefore, it can be easily send to another com-
puter, device, or server easily. Size of yellow pages is configurable. If there is no space to write
in yellow pages, it overwrites the oldest advertisement/message by using newest advertisemen-
t/message. Once manager provides needed resource for particular node, the advertisement that
is published by that node, will be deleted.
The mailing system is implemented to collect information from nodes continuously. It can
be considered as offline messaging system as well. It is called offline system since the messages
between node and manager are not sent immediately and even there is not any connectivity
between them, it saved and sent it later once get the connectivity again. A node creates a mail
and registers it into its mailbox. Manager checks mailbox of nodes regularly and collects these
mail from them. There are two types of mail that can be registered by nodes. These are infor-
mation and publish ad mails. Nodes register information mail periodically. Information mail
contains seven different information about particular node. These are needed, available and
total resources, number of ready and total task, battery percentage and timestamps. Mailbox
is implemented as array of structure as well so there is a configurable limit. A node overwrites
the oldest information mail with new one if there is no space in the mailbox. However, pub-
lish ad mails are sent immediately since they contains information about the need of the node.
This mailing system is implemented together with the yellow pages to share information and
resource along the network with the low power consumption. The structure of mailing system
and yellow pages can be seen at the figure 8.5.
This service handles with radio callbacks. This particular operating system has two different
Manager
Node 1
Memory Operating System
Registered info mail
INFO Info Mail IOT INFO

İnfo mails IOT Service İnfo mail
MAILBOX
MAILBOX Service
Publish ad
Operating System Memory
Yellow Registered info mail

advertisements
Pages
Info Mail IOT INFO
İnfo mail
Service MAILBOX
Publish ad
Operating System Memory
Node 2
Figure 8.5: Structure of Mailing System and Yellow Pages
point of view as it was explained before. One of them is user point which can be thought as
writing firmware without dealing with any communication library and protocol. The IOT ser-
vice I is implemented for that purpose. This service handles with all wireless communication
including handling with radio module callback. Call-back function for radio module is defined
and implemented into this service. Radio library sends only radio state to higher-level callback.
A buffer is implemented as well to write received data. According to radio state and device state,
service extract relevant data from the message. The purpose of handling radio callback in this
service is to not block other kernel services or OS exceptions or other hardware interrupts with
wireless communication. Since this is an embedded operating system, it needs catch if there is
a hard deadline to catch. In addition, the IOT service simply can be degraded by giving lower
priority if user needs to execute more important task.
typedef enum{
IOT_SERVICE_RESOURCE_TYPE_TIMER ,
IOT_SERVICE_RESOURCE_TYPE_LED ,
IOT_SERVICE_RESOURCE_TYPE_ACTUATOR,
IOT_SERVICE_RESOURCE_TYPE_COMPUTATION,
IOT_SERVICE_RESOURCE_SENSOR_TEMPERATURE,
IOT_SERVICE_RESOURCE_SENSOR_HUMIDITY,
IOT_SERVICE_RESOURCE_SENSOR_PRESSURE,
IOT_SERVICE_RESOURCE_SENSOR_PROXIMITY,
} iot_service_resource_type_t ;
Listing 8.1: IOT node resources
A node may have various type of resources. These type of resource is implemented as a spe-
cial type called iot_service_resource_type_t and it can be seen at the listing 8.1. For differ-
ent kind of applications, node would have different modules and might need different types of
resources. For instance, a group of node inside of a cluster may have the task to measure tem-
perature while other group measures humidity and assume that they have both sensor module
mounted on the devices. Nodes use only one sensor at the time(humidity or temperature) to
reduce power consumption. If one of the nodes has failure in its temperature sensor, it can pub-
lish an advertisement about the need of temperature sensor and manager can control and use
the temperature sensor on the other node on behalf of the needed node.
typedef enum{
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_LED_ON,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_LED_OFF,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_LED_TOGGLE,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_ACTUATOR_ON,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_ACTUATOR_OFF,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_ACTUATOR_TOGGLE,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_SENSOR_READ,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_SENSOR_WRITE,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_TIMER_START,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_TIMER_STOP,
IOT_SERVICE_RESOURCE_SLAVERY_COMMANDS_TIMER_SET,
} iot_service_resource_slavery_commands ;
Listing 8.2: IOT node resources
There are many ways to control node by manager inside of a cluster. This command called
slavery command in the particular protocol. Basically, every node opens radio module in receive
mode periodically to check whether there is any request from manager or not. At the figure 8.2,
the type of slavery command can be seen. Sensors, timers, actuators and LEDs can be control
easily by manager using these commands. More functionalities are considered and discussed in
the future work section later on this paper.
IOT service task has also provides an API to user to use the IOT service. User can start and
stop IOT service using this API. In addition to that, API tracks the number of the send packet and
received packets. IOT service registers every node in the cluster inside of one list. API provides
information about each node and that list to user as well. Moreover, user can interfere IOT
manager task and can start slavery with the entire node in the cluster regardless their workload.
However, it is not recommended to interfere most of the time, if application of the other node is
not known well.
8.3 Security Issue
The security issue of large sensor networks and IOT devices have been becoming more impor-
tant day by day. In addition to that, according to estimations each person in the world will have
had six devices. Even now all around us is surrounded by personal devices, smart buildings,
even cities and factories. We as developers use these devices to measure our heartbeat, temper-
ature, location, blood pressure. We make everything accessible. Even our most sensitive and
private data. For instance when you set your Wi-Fi-enabled devices, you add something to your
local area network without proper security. That means you put a device after your security or
firewall. That could be used as a key to your more private data even to your computer.
The risks are obvious as we discussed earlier in this paper. Rogue/Mock/Invalid firmwares,
invisible backdoors, eavesdropping, having malware in upgrade. In addition, devices called bot-
nets are already a major threat. Number of attack against routers, smart TVs, game consoles are
increasing. Limited encryption capabilities of device one of the main issue. In addition, upgrad-
able software is the key concept. It is hard to find backdoors or bugs without putting particular
product to the field. Linux, Windows, Android and IOS all the other big OS they always fix their
bugs and backdoors with patches.
Security should not be thought of as an add-on to a device, but rather as integral to the
device’s reliable functioning. Software security controls need to be introduced at the operat-
ing system level, take advantage of the hardware security capabilities now entering the market,
and extend up through the device stack to continuously maintain the trusted computing base.
Building security in at the OS level takes the onus off device designers and developers to con-
figure systems to mitigate threats and ensure their platforms are safe. Therefore, it can be easily
seen that specific OS for IOT is needed. Thus, main purpose of this service and protocol to pro-
pose a new way of thinking to IOT and large sensor networks and security issue.
As it was explained before, in this particular protocol, each node registered to the cluster
uses a different channel to communicate with the manager. The information about which node
communicate in which channel is only stored in manager. Therefore even node is accessed
by unauthorized user, there is no way to get the knowledge of the channel information regard-
ing to every node. On the other hand, the node needs to keep creating new information mails
periodically. Therefore, any missing or suspicious information coming from node side can be
considered as threat and node can be banished from the cluster.
One of the component that also provide security in operating system level is communica-
tion layer. Communication layer provides abstraction between radio module and application.
Therefore even unauthorized user somehow access to a node, it can only use radio module with
limited functionality provided by API. As it was discussed before, it can only publish advertise-
ment about needed resource and it needs to be approved by manager too. Thus, there is no way
to access manager or radio module except this.
IOT service is defined as user tasks in the scheduler. Its only privilege than regular user tasks
is having higher priority than all the other user task. Therefore IOT service can use some spe-
cific functionalities provided by kernel but cannot be run in kernel mode or privilege. Moreover,
other functionalities of operating system, exception and interrupt handling has different prior-
ities than scheduler’s task. Thus, kernel and other components of particular operating system
still keep working even IOT service controlled by unauthorized access.
In this particular network structure, manager has many rights on network and nodes. It can
collect information, use their hardware and can assign task to do. On the other hand, node
device type has very limited right or functionalities over the network and devices. A node can
only access to the manager using assigned channel. Manager device decide to accept the node
to the cluster and registered it, or send rejection message to the node. Moreover, manager can
assess all the collected information from nodes and can perform a health check of entire net-
work or particular node. Since any suspicious act of node will degrade the network health point,
manager can shut down the entire cluster or send emergency signal to another trusted device.
As it was mentioned before, a registration system is implemented in this network structure.
After booting-up of particular node, it starts to send registration message in fixed frequency
called registration frequency, which is 2.44GHz. It performs that task periodically to reduce
the power consumption. Same functionality is also performed by manager device. It periodi-
cally checks whether there is new device try to register in registration frequency or not. If there
is a node which is trying to register, it checks the protocol and extract the message. After the
protocol check, manager registers and stores the information about particular node and send
an acknowledge message about registration process. If it is successful , particular node stops
searching new manager, starts performing other user tasks while creating information messages
periodically. This registration process make registration of mock/invalid or unauthorized device
harder. Moreover, any suspicious act and timeout of node is being observed by manager and
these kind of acts can be considered as invalidation of particular node, thus, it can be banished
from cluster. The flow chart of the registration process for manager and node side can be seen
at the figure 8.6.
Manager Device Node Device
Peroidic event Peroidic event

occurs occurs
Send Registration
Is there any Start timer for Start timer for
N Message from
node? next period next period
fixed freq
Did timeout N
Is protocol happen?
N
correct?
Listen fixed freq
until timeout
Send unsuccesful
Y
registration
Y
acknowledgment
Is there any N
Is there any space N
ack?
N
for a new node?
Y
Is it succesful
registiration?
Send succesful
registration y
acknowledgment
Store Manager
information
Register the node

Execute Other Tasks
Figure 8.6: Flow Chart of Registration Process

Chapter 9
Tests and Results
After the every software development, exhausting tests must be done to validate the function-
ality of the system. Verification and validation is the process of checking a software system
meets specification and perform expected functionality. Software verification is ensuring that
the product has been built according to the requirements and design specifications. On the
other hand, verification in software development checks that functionally of devices whether as
it is expected or not.
In addition, while testing a particular software of software related product, terms should be
used in order to explain the type of the problem. Therefore, from test point of view, fault means
wrong or missing function in the code. The manifestation of a fault during execution is called
failure and malfunction of software or device can be described as according to its specification
the system does not meet its specified functionality. In this particular section, tests and test
scenarios for particular operating system are explained in detail. In addition to that, results,
which were obtaining after the tests are presented.
In order to test the operating system in real world, embedded devices (a circuit-board or
development kit) were needed to implement particular project in it. For that purpose, two types
of devices have been used provided by vendor company Nordic Semiconductor. As it can be
seen from figure 9.1, these are NRF518222 Development Kit and NRF51822 Dongle.
The selected SOC NRF51822 has two type of development or evaluation boards provided by
vendor. One of them, which is used in the test as node device, is nRF51 Dongle. It is a low-cost,
versatile USB development dongle for Bluetooth Smart, ANT and 2.4GHz proprietary applica-
99
CHAPTER 9. TESTS AND RESULTS 100
Figure 9.1: NRF518222 Development Kit and NRF51822 Dongle
tions. The kit provides I/O interface with 6 solder pads and has programmable RGB LED and
programmable button for user applications. The other development kit that is used in this par-
ticular project is NRF51DK. It is used as manager node during the tests and it is a low-cost,
versatile single-board development kit for Bluetooth Smart, ANT and 2.4GHz proprietary ap-
plications. The kit provides I/O interface via connectors and has 4 programmable LEDs and 4
programmable button for user applications. Both of the devices provide Program/Debug op-
tions with Segger J-Link OB for standard tool-chain.
Basic functionality of operating system has been tested and verified by using NRF51822DK
device. After achieving the satisfactory result from kernel functions and hardware drivers, IOT
service layer and radio infrastructure had been started to develop. In order to test radio library
and functionalities, a system formed by a NRF51822 DK board and a NRF51822 board, was cre-
ated and ping-pong test was done between two devices. Afterwards, a test has been formed by
same devices but this time with IOT service layer and new radio protocol implemented. After
we ensured that IOT service layer performs communication between devices and all the func-
tionalities related with communication is working, resource and information sharing part of the
layer was started to develop.
Information
about entire
Cluster
Information Mail,
Information Mail, Adveritsement Adveritsement,Slavery Ack
Node 3 Manager Node1
(nRF51 Dongle) Mail request (The nRF51822 Evaluation kit) (The nRF51822 Evaluation kit)
Slavery Command, Mail request
Information Mail,
Adveritsement
Mail request
Node 2
(nRF51 Dongle)
Figure 9.2: Structure of Test Scenario
As it is illustrated in figure 9.2, this cluster was formed to test all the system functionality
and timing specification for IOT service layer. In this test scenario1 , a user program has been
developed in the manager device, to communicate with personal computers and send the in-
formation sharing inside of the cluster. In addition, same program also was controlling Node1
LEDs by using IOT service API. Computational tasks were implemented all the devices in the
cluster including manager to see the effect of multitasking over IOT service layer. In this test all
the nodes issued mails in the memory and manager collected those mail periodically and infor-
mation about each node was seen on from personal computer serial terminal. After the tests, it
has seen that all the functionalities are implemented such as resource and information sharing,
task execution, hard-drivers etc. In addition to that, size of the code and execution time of par-
ticular IOT service layer are obtain as a results of the test. Particular results are shown in table
9.1 and 9.2 .
1
The example cluster which is formed by three nodes and one manager, is also submitted with this thesis
Memory Partition
Device FLASH RAM IOT SERVICE
Type SIZE SIZE LAYER SIZE
Node Device 6.43kB 7.58kB 1.375kB
Manager Device 6.95kB 8.24kB 1.785kB
Table 9.1: Size of the Embedded Operating System
Communication Type
Device 1-1 Cluster
Type Communication (4 Devices)
Node Device 207.5ms 210ms
Manager Device 267.5ms 866ms
Table 9.2: Execution Time of IOT Service Layer

Chapter 10
Discussion, Future Work, and Conclusion
10.1 Discussion
The main motivation of the thesis was to design a simple kernel and a small embedded oper-
ating system particularly for low-power resource-constrained devices with the additional layer
called IOT layer and make development easier for the trending concept. To achieve this, some
research had to be done. However, before designing such a system, requirements fo the system
should have been decided. Hence, detailed investigation about different embedded operating
systems and distributed system architectures was done. Since the competition in the embed-
ded OS market fierce, proposed architecture should introduce new features to that area. Main
focus while designing the operating system was to keep the simplicity of the design as much as
possible while providing different functionalities to users with the operating system.
On the other hand, the footprint of the OS was also the other main concern. Therefore,
creating own kernel architecture was seemed quite promising. This task was indeed time con-
suming but eventually a “microscopic” kernel was designed which satisfies our needs and re-
quirements. Moreover, a very simple and new protocol for wireless communication is desired
to be implemented to enable wireless communication and information sharing. This protocol
is implemented together with the IOT service layer which is the novelty of the project. Relevant
tests about the functionality and performance have been done and the results are shown earlier
in this thesis.
As it was mentioned before, this operating system is designed specifically for the low cost,
103
CHAPTER 10. DISCUSSION, FUTURE WORK, AND CONCLUSION 104
low power and resource-constraint embedded devices. Currently in the market many differ-
ent embedded devices and CPU architectures can be found. Especially single-board computers
started to dominate the market. This device type is a complete computer built on single board
with input/outputs, memory and microprocessor. Moreover, single-board computers such as
raspberry pi or beagleboard often have features like graphical-user interface, wireless shields
to enable wireless communication and other popular communication protocols such as USB,
HDMI, etc. Mostly they are used for every kind of applications and educational purposes. More-
over, most of them contains large processor with lots of features and large memories.
During the development of this particular project using these single-board computers would
be easier and faster. However, these devices are not good fit for sensor nodes or low-power IOT
devices. Firstly, these devices consumes more power than small microcontrollers since they
have many module in it, thus it is not convenient to use those devices with battery and most of
the IOT and sensor network application requires low-power consumption and mobility. More-
over, during our research we have seen that, most of the challenges in those devices are going
around industry and user needs. Therefore, we could not find academic challenges which mo-
tivate us to research on single-board computers.
In contrast, resource-constrained devices have many challenges to overcome since it is al-
ways hard to perform complicated tasks with resource-constrained devices. In addition to this,
with the growing IOT and sensor network concept, many new problems and challenges have
appeared such as security, efficiency, power consumption and functionality.
In this project, we particularly focused on the task execution and scheduling in the embed-
ded operating system. Our target architecture for development was low-power single core pro-
cessors especially ARM Cortex-M series and as it was explained before a cortex M0 architecture
has been used as a target during the development. The purposed kernel architecture can sup-
port multitasking. It can switch between task and run them according to their priority.
However, using memory management units and algorithms, multi-thread behavior could
have been achieved. With this method two task can be run concurrently. However, this method
would cause performance problems in entire system .Moreover, it might affect the execution of
real-time task. Since particular operating system is designed for embedded system and devices,
real-time compatibility is an important feature.
On the other hand, a true concurrent kernel could have been developed and implemented
on the multi-core processor architecture. However, multi-core systems are completely different
topic. Developing a kernel for multicore processor would be very time consuming. In addition
to this, developing such an operating system has its own academic and technical challenges.
Therefore, the project related with a multicore processor would not be related with our main
focus which is a small and efficient operating system for sensor and IOT device network.
In order to achieve small footprint for the operating system in this project, optimization
must be done in the most of the operating system components. Usually, one of the largest part
of the operating system is OS services. Despite their large size, OS services are the component
which makes operating systems out of piece of software and kernel functions. Not only commer-
cial operating systems for personal computer such as Windows, Linux or Macintosh but also in
embedded OSes such as FreeRTOS, TinyOS, Contiki. There are many OS services running in the
background while the user application is running.
The number and types of OS servces in embedded operating system are very controver-
sial topics. During our research, we have seen that most of the embedded devices uses only
task scheduling and communication drivers in sensor networks and IOT devices. Therefore,
it has been decided that particular operating system would provide task scheduling/execution
and communication drivers as an operating system services. All the other services that can be
needed, depends on application such as file management can be added by the user. In addition
to that, this operating system has been designed to provide an infrastructure to enable informa-
tion and resource sharing across the sensor or IOT device network. Because of its user-friendly
design, system can be extended easily from user layer for more demanding applications.
The kernel is the most important part of the operating system and can be imagined as the
heart of every operating system. The kernel design has been being very popular topic both
in the industry and academy. Different operating systems use different kernels. Monolithic,
micro, hybrid, nano and exo kernels are the widely used ones in the industrial and academic
projects. Moreover, there was a big debate called Tanenbaum-Torvald debate about microkernel
and monolithic kernel architecture.
All these kernel types have different advantages, challenges and disadvantages. However, a
strict structure like monolith or microkernel is not a suitable architecture for embedded appli-
cations. These structures are very good in a system with powerful CPU and many resources to
manage. Nevertheless, most of the non-constrained embedded systems in sensor and IOT net-
works have low profile CPU and limited resource to manage. Therefore, to implement such an
architecture in this project, would be definitely in vain.
On the other hand, efficiency of the task execution appeared to be more important than
number of services provided by OS according to background research. In addition to hat, we
believe that a new kernel design is needed for the resource-constrained device in the sensor and
IOT device networks. Thus, a new kernel architecture has been proposed and implemented in
this thesis.
A priority driven scheduler is implemented inside of proposed kernel architecture. This type
of schedulers are easy to implement. The scheduler does not need to know the information
on the release times and execution times of the jobs. In addition, the run-time overhead due
to maintaining a priority queue of ready jobs can be made small. The context switching does
not occur often in this type of systems thus also overhead caused by context switching does
not cause any problem. The priority-driven algorithms are on-line scheduling algorithms. The
scheduler makes decision without having information about jobs released in the future. More-
over, on-line scheduling is very suitable option for a system which future workload is unpre-
dictable such as embedded systems.
In contrast, priority-driven schedulers are non-deterministic, thus the timing behavior of
the system is not predictable. In addition to that, it is difficult to validate that all jobs sched-
uled in a priority-driven manner meet their deadlines when the jobs parameters vary. There are
better scheduling algorithms for real-time systems such as earliest-deadline first, value-based
scheduling or deadline monotonic priority algorithms. Moreover, priority-ceiling algorithms
could have been implemented with dynamic scheduler to avoid priority inversion and dead-
locks during the execution of the real-time task. However, operating system needs to control
the tasks all the time and know everything about the tasks in these algorithms. That means a
bigger task library thereby a bigger kernel and bigger footprint of the operating system. Ad-
ditionally, this would be completely opposite of the main idea of the particular project that is
simplicity.
After research on the sensor networks and the IOT concept, it is revealed that intelligent
networks for these kind of networks could be the solution to overcome the challenges. In order
to achieve that purpose, each network should have conscious to investigate what is going on the
network and act accordingly. This could be possible with information and resource sharing.
As it was discussed before, these kind of networks are full of resource-constrained devices.
Since achieving information sharing across network would increase the communication rate
per device per second, this would lead increasing of power consumption. However, one of our
main motivation for this project was a specifically designed operating system and OS service.
This could be successfully enable resource sharing and keep the power consumption as low as
possible.
However, during the development, we have seen that not only the OS and OS services but
also hardware drivers and communication infrastructure should be designed specifically. To
implement communication infrastructure for particular IOT service, Bluetooth protocol could
have been used. It is a compatible, low-power protocol and all the libraries are provided by
vendors. However, the Bluetooth stack is too large to implement on such an operating system.
Bluetooth stack must be optimized or reduced. Even it would not have had resource-sharing
feature. Therefore, although it is indeed time consuming to develop a specific protocol, a li-
brary and a service for this purpose, a simple structure was purposed and implemented into the
operating system in this thesis.
While developing the project, some decision must have been made to increase develop-
ment efficiency. Since this project includes complete development of an operating system from
scratch including device drivers, software platform and the compiler that are used in the de-
velopment are very important. After deep investigation about development environments for
targeted platform, in this particular project, Keil uVision5 software development platform has
been decided to use. This environment specifically developed for wide range of Cortex-M, and
Cortex-R based microcontroller devices. In addition, it provides the µVision IDE/Debugger,
ARM C/C++ Compiler, and essential middleware components.
As a compiler, Keil provides a complete tool chain called the ARM compiler toolchain. This
includes following components:
• The ARM C/C++ Compiler (armcc)

• Microlib
• The ARM Macro Assembler (armasm)
• The ARM Linker (armLink)
• ARM Utilities (Librarian and FromELF)
With the help of these tools developers can easily write applications for the ARM family mi-
crocontrollers in C or C++. Moreover, after compilation this tools ensures to provide speed and
efficient of assembly language. The tool chain translate C/C++ files into relocatable object mod-
ules which contain full symbolic information for debugging with the µVision Debugger or an
in-circuit emulator. Further more the compiler generates a listing file which contains symbol
table and cross-reference information.
These development tools for the ARM family of microcontrollers allow you to write ARM
applications in C or C++ that, once compiled, have the efficiency and speed of assembly lan-
guage.The ARM Compiler toolchain translates C/C++ source files into relocatable object mod-
ules which contain full symbolic information for debugging with the µVision Debugger or an
in-circuit emulator. In addition to the object file, the compiler generates a listing file which may
optionally include symbol table and cross-reference information.
As it was discussed before cortex-M0 processor architecture has been decided as targeted
architecture. It is because this particular architecture is a very low gate count, highly energy
efficient processor that is intended for microcontroller and deeply embedded applications that
require an area optimized processor. In addition to that its power consumption is lower than
other competitor architectures.
The configurable, multistage, 32-bit RISC processor ARM Cortex-M0 processor consumes
small silicon area and low power and also minimal code footprint. Due to all these, developers
can achieve higher performance than old 8-bit micro controllers. Therefore, it is also very suit-
able architecture for embedded system learners and developers in academy. In addition to that,
this particular project is also based on wireless communication and designing a new infrastruc-
ture for sensor networks and IOT devices. Therefore, this particular architecture was the most
suitable architecture for this project since it is specifically optimized connectivity to support low
power protocols such as as Bluetooth Low Energy (BLE), IEEE 802.15 and Z-wave.
Detailed investigation has been done to choose most suitable microcontroller/SOC and its
development kit to implement the software and test it. NRF51 series from Nordic Semiconduc-
tor is the most suitable for this project in price and performance manner. NRF51 is a system-
on-chip with with a Cortex M0 and a BLE radio chip all in one. With provided libraries it is easy
to learn and develop software. NRF51822 has been selected among NRF51 series. The particular
SOC has two type of development boards provided by vendor. One of them, which is used in the
test as node device, is nRF51 Dongle. The other development kit that is used in this particular
project is NRF51DK and it is used as manager device.
In this particular project a kernel designed has been proposed for embedded devices specif-
ically sensor and IOT networks. The kernel architecture is implemented together with a new
radio protocol and library into an embedded operating system which has specific IOT service.
All the software which was develop during the project is open source and anyone from indus-
try or academy can access and use it easily. In addition, this particular paper would lead the
way those who would use this embedded operating system. As it was explained before purpose
of proposed kernel architecture is to create user-friendly environment. Moreover that this par-
ticular operating system is fully functional currently. A group of node has been controlled by
manager device and a cluster has been created as it was mentioned in “test and results” section.
On the other hand, the particular operating system is open to develop further by those who
would like to continue this project in the academy. Although, future works is explained in de-
tail in the next section, there are many other ideas to extend this project. This system is not only
introduced as user-friendly but also it is developer friendly because of its two dimensional archi-
tecture. Furthermore, this software can be turned out to be an industrial or commercial product
with few changes. Although it is a master project and developed with an amateur spirit, after
necessary tests, validation and improvements it would be a competitive product in the market.
10.2 Future Work
The software development in this project was done from scratch. It was tested and the result of it
is ensured that the system functionality works as it was expected. This project also proved that
our concept ideas on sensor and IOT network could be implemented and used in embedded
devices. However, during the background research and development of the project, new ideas
have appeared about the project and further development of the project. Since, the time was
limited while developing this project, this ideas could not be implemented into this project.
Therefore in this section, recommendations for the possible extension of the project are given.
These recommendations can be classified as short, medium and long-term future work.
For short term;
• Although this particular operating system designed to achieve low-power consumption,

no power modes is implemented for different type of application. Therefore different
power modes with different specification can be added and the power management can
be optimized in future works.
• One of the concern while developing IOT layer and communication infrastructure that
lies behind it, was security. In architectural design of this particular layer and abstraction
of communication library helps user to avoid security issues. In addition to that, most
of the microcontrollers have AES encryption modules. Communication between nodes
and manager can be encrypted by using this modules and more secure platform can be
developed.
• Although, many hardware libraries were developed during the project, some of the li-
braries, which is commonly used in embedded application, could not be implemented
due to limited time. Libraries such as flash memory, SPI, I2C can be developed and inte-
grated easily.
• This system and communication infrastructure is specifically developed for sensor and
IOT device networks. Particular network infrastructure can be configured to use specific
application such as server application, distributed system applications or robotics appli-
cations.
For medium term;
• As it was explained before, manager node contains all the information about the clus-
ter and control resource of nodes as its own resource. For end user a user interface can
developed for Windows or Linux via USB port since all of the personal computer has USB
ports and communication between PC and manager node is already tested by using UART
library with converter module from UART to USB.
• Information and resource sharing features are implemented and tested in this project.
However, assign task sharing and using other nodes computational power could not be
implemented due to limited time.
• None of the memory management concepts such as paging or caching were not desired
to developed to not lose the focus of the project. However, for high profile CPUs, memory
management support can be implemented to increase the total system performance.
• At the beginning of the project, security layer between IOT service layer and application
layer was thought to be one of the feature of the embedded operating system. However,
it is cancelled to implement after the careful consideration about workload and time. A
registration and validation system can be implemented for user application to user IOT
service. With that kind of system, unauthorized accessed node can isolated from the clus-
ter and possible attack can be prevented.
• In addition, a tool for yellow pages can implemented inside of manager node which can
work as database. This data can be compressed while storing in database and this would
lead smaller footprint as well as having more functionality on yellow pages.
For long term;
• This particular project is designed to implement many different processor architectures.

For that purpose, device specific components such as hard drivers and portable kernel
functions have been abstracted. Therefore, to provide support for different processor ar-
chitecture would be easy. However, since there are many different processor architectures
in the embedded system market, that task indeed would be time consuming.
• Producing multicore processors with low power consumption feature is trending concept
in integrated circuit industry. Despite the technical challenges of multicore system, par-
ticular kernel can be extended to support these kind of architectures.
• During the background research, problems and challenges in the networks caused by
number of device and capabilities of devices have been revealed as most difficult chal-
lenges and problems to overcome. However, we proposed earlier in this paper, these chal-
lenges and problems could be overcome implementing machine learning or AI algorithms
to network infrastructure and service. In addition to that, since particular project achieves
wireless communication thus, it could be used to implement those algorithms to sensor
or IOT device networks.
• The particular operating system provides real-time compatibility for real-time applica-
tions. However, since a priority-driven scheduler is implemented into the kernel, it may
degrade the performance or may cause some problems in hard real-time systems. In or-
der to solve this problem, different scheduling algorithms can be implemented. Yet, task
library needs to be extended to implement those algorithms.
• This particular operating system has been developed for engineers and developers. There-
fore, end-user cannot use particular system without having any knowledge about com-
puter science, operating systems and electronics. However, a graphical-user interface can
be developed as it is in end-user operating systems. With a simple GUI, user could see the
network elements of particular network and manipulate network structure and devices in
the cluster.
10.3 Conclusion
The software development for embedded devices such as low-power and small microcontroller
has been significant topic for last 10-15 years. Many research have been going on both in the
academy and industry about efficiency and functionality of embedded devices. Therefore, many
different operating system optimized specifically for embedded devices, have been developed.
Together with wireless connectivity, embedded devices have started to appear everywhere in
our daily life. Nowadays, it is an important concept for not only in the academy and industry
but also in regular life. Therefore, embedded devices will have been important concept for next
years. However, since wearable technologies and Bluetooth low energy devices have started to
be used in the most private and crucial applications, the importance of failures in these devices
and networks have increased. On the other hand, the old technology has started to be thrown
away such as old 8-bit microcontrollers because of the trending SOCs with wireless communica-
tion modules. These devices should be reused to prevent the world from being microelectronic
garbage. Moreover, because of the funding problems, most of the universities in non-developed
countries cannot afford embedded operating systems to teach the basics of computer science
and electronics.
Although, the developments in IC technology and software tools has provided improved
functionalities to the small microcontrollers, it also has started to cause new problems. There-
fore, both in the academy and industry a new point of view for embedded operating systems is
very important. In this thesis, we have proposed ideas and solutions and we have shown that
these problems could be solved even in limited-time with limited labour. After the detailed re-
search and dedicated software development, an embedded operating system was designed in
the light of new ideas. Moreover, during this project, problems and challenges of the new types
of networks which are formed by hundreds of sensor and IOT devices are revealed. In order
to solve these problems, a radio library were specifically developed for this project. Moreover,
a wireless communication protocol was proposed and implemented. Furthermore, to provide
improved functionality such as information and resource sharing for elements of the networks,
a new operating system service and layer called IOT service layer was developed and imple-
mented.
In order to develop such an embedded operating system, some specific tasks must be de-
fined and done. First task was to develop required hardware drivers for the targeted platforms.
Although, the vendor company of the targeted platform provides libraries for the hardware,
they must have been developed from scratch to have efficient, optimized and suitable hard-
ware drivers for the operating system. Moreover, one of the important feature of the project was
the proposed kernel architecture. This kernel is designed to achieve not only small size but also
to be a real-time kernel. In addition, the OS services must be optimized and implemented on
the operating system to make the operating system functional. However, the most difficult task
in this project was developing the specific radio library, a protocol and a service, which depends
on a new communication infrastructure. Detailed research about wireless communication and
distributed system has been done to achieve these tasks. Every software project must be tested
and the proposed features and functionalities must be verified. Therefore, a test scenario has
been created and the entire system features and functionalities has been tested by using devel-
opment kits provided by the vendor.
This operating system has features and characteristics, which are different from other em-
bedded operating systems in the market. The system has very small footprint. It can be imple-
mented even on modern small microcontrollers in the market without need an external mem-
ory. Simplicity of the system provides users an efficient task execution. Moreover, the proposed
two-dimensional kernel architecture gives users a simplified point of view for the development
thus; it leads the system to be more user-friendly. Last but not least, the operating system and
communication protocol specifically designed for the sensors and IOT device networks. It can
be configured easily to use in such networks.
In this particular project, a new architecture for embedded operating systems has been pro-
posed and a small embedded operating system with distributed system functionality has been
designed and implemented. In addition to this, the operating system was implemented on dif-
ferent resource-constrained devices with same processor architecture and detailed tests about
system have been done. Currently, all the features and requirements, which were mentioned in
the introduction part of this thesis, have been met and the embedded operating system is ready
to use. Moreover, during the development of the project, detailed background research were
done about operating systems, distributed systems and wireless communication and explained
in this thesis. Therefore, this paper can be considered as a resource in these topics as well as the
user-manual of the embedded operating system. Furthermore, a conference paper about the
operating system particularly IOT service layer is being written.
Lastly, the current developments in the technology are increasing the importance of the em-
bedded systems and resource-constrained systems in our daily life. The application field of the
embedded systems has been growing wider thus; the stored data and handling with system fail-
ures have been becoming more important as well. Therefore, the operating systems or network
protocols, which are used in these applications, should be enhanced. More efficient and secure
systems are needed. Since the vast wireless sensors and IOT device networks surround us all
over the world, even the slight improvements in the operating systems or network architectures
will produce significant results.

Chapter 11
Bibliography
[1] ITU, "Internet of Things Global Standars Initative," [Online]. Available: http://www.itu.int/en/ITU-
T/gsi/iot/Pages/default.aspx. [Accessed 1 April 2016].
[2] D. Gordon, "The Trojan Room Coffee Machine," University of Cambridge, [Online]. Avail-
able: http://www.cl.cam.ac.uk/coffee/coffee.html. [Accessed 1 April 2016].
[3] A. Wood, "theguardian," [Online]. Available: http://www.theguardian.com/media-network

/2015/mar/31/the-internet-of-things-is-revolutionising-our-lives-but-standards-are-a-must.
[Accessed 1 April 2016].
[4] J. Apcar, "IP Routing in Smart Object Networks and the Internet of Things," Melbourne ,
2012.
[5] UN Economic and Social Affairs, "World Population to 2300," Cisco IBSG projections, 2004.
[6] P. Bennett, "EE Times," [Online]. Available: http://www.eetimes.com/document.asp?doc_id=

1276973. [Accessed 2 April 2016].
[7] H. Mario, T. Pentek and O. Boris, "Design Principles for Industry 4.0 Scenarious: A literature
Review," Technische Universität Dortmund, 2015.
[8] Z. Technologies, "Variable global enterprise Internet of Things decision makers," 2014.
[9] S. P. M, "Embedded Operating Systems for Real-Time Applications," M.Tech. credit seminar
report, Bombay, 2002.
116
CHAPTER 11. BIBLIOGRAPHY 117
[10] A. S. Tanenbaum, Modern Operating Systems, Amsterdam: Pearson Education Interna-

tional, 2009.
[11] I. C. Bertolotti, "Real-Time Embedded Operating Systems: Standards and Perspectives," in

Embedded Systems Handbook, Taylor and Francis Group, LLC, pp. 4-10.
[12] BrokenThorn Entertainment„ "Operating Systems Development - Kernel: Basic Concepts

Part 2," BrokenThorn Entertainment, [Online]. Available: http://www.brokenthorn.com/
Resources/OSDev13.html. [Accessed 06 05 2016].
[13] S. Poudel, ""Operating System Structure"Science HQ," Ed. Rod Pierce, 2013 February 18.
[Online]. Available: http://www.brokenthorn.com/Resources/OSDev13.html. [Accessed
6 May 2016].
[14] A. Holt and H. Chi-Yu, Embedded Operating Systems, Bristol: Springer, 2014.
[15] G. Coulouris, Distributed Systems Concept and Design, Boston: Pearson, 2012.
[16] J. WEIJIA, Distributed Network Systems from Concepts to Implementations, Boston: Springer,
2005.
[17] D. Kinneryd and A. Mäkitalo, "Distributed applications: A journey to platform indepen-

dency," Luleå University of Technology, Luleå , 2007.
[18] tutorialspoint.com, "Tutorials Point," tutorialspoint.com, [Online]. Available: www.tutorials

point.com/software_architecture_design/distributed_architecture.htm. [Accessed
8 May 2016].
[19] Micro Focus, "Networking Primer," Novell, [Online]. Available: https://www.novell.com/

info/primer/prim05.html. [Accessed 8 May 2016].
[20] H. Karl and W. Andreas, Protocols and Architectures for wireless sensor networks, Chich-
ester: John Wiley and Sons, 2005.
[21] Magna Design Net, "Digital Modulation," Magna Design Net, 2014. [Online]. Available:
http://www.magnadesignnet.com/en/booth/technote/ofdm/page2.php. [Accessed 8 May
2016].
[22] Maxim Integrated, "An Introduction to Spread-Spectrum Communications," Maxim Inte-

grated, 18 February 2003. [Online]. Available: https://www.maximintegrated.com/en/app-
notes/index.mvp/id/1890. [Accessed 2016 May 8].
[23] RS Components, "11 Internet of Things (IoT) Protocols You Need to Know About," RS Com-
ponents, 2015. [Online]. Available: http://www.rs-online.com/designspark/electronics/
knowledge-item/eleven-internet-of-things-iot-protocols-you-need-to-know-about. [Ac-
cessed 2016 May 8].
[24] C. Links, "Wireless Communication Standards for the Internet of Things," Green Peak
Technologies, 2015.
[25] C. Svec, "The Architecture f Open Source Applications Volume II," Creative Commons:
Attribution, 2012.
[26] S. Kolesnik, "Comparing microcontroller real-time operating systems," embedded.com,

2013.
[27] A. Dunkels, "Full TCP/IP for 8-Bit Architectures," in The First International Conference on
Mobile Systems, Applications, and Services, San Francisco, 2003.
[28] A. Dunkels, O. Schimdt, T. Voigt and A. Muneeb, "Protothreads: Simplifying Event-Driven

Programming of Memory-Constrained Embedded Systems," in SenSys’06, Boulder, 2006.
[29] B. Ganesh, "Architectural Support for Embedded Operating Systems," University of Mary-
land„ Maryland, 2002.
[30] H. Galzner and M. Hartmann, "TinyOS, an Embedded Operating," Faculty of Informatics,

TU Wien, Wien, 2012.
[31] V. R. Aroca and G. Caurin, " A Real Time Operating Systems (RTOS) Comparison," Univer-
sity of Sau Paulo, Sao Paulo, 2009.
[32] C. L. Wang, B. Yao, Y. Yang and Z. Zhu, "A Survey of Embedded Operating System," Univer-
sity of California, San Diego, 2001.
[33] Nordic Semiconductor, "nRF51822," Nordic Semiconductor, [Online]. Available: https://

www.nordicsemi.com/eng/Products/Bluetooth-Smart-Bluetooth-low-energy/nRF51822. [Ac-
cessed 10 May 2016].
[34] Nordic Semiconductor, "nRF51822 Product Specification v3.1," Nordic Semiconductor,

Trondheim, 2014.
[35] Nordic Semiconductor, "nRF51 Series Reference Manual," Nordic Semiconductor, Trond-
heim, 2014.
[36] ARM Ltd., "Cortex-M0 Processor," ARM Ltd., 2012. [Online]. [Accessed 10 May 2016].
[37] J. Yiu, The Definitive Guide to ARM Cortex-M0 and Cortex-M0+ Processors, Elsevier, 2015.
[38] S. Rollins, "Linked Lists," University of San Francisco, 01 October 2007. [Online]. Available:
http://www.cs.usfca.edu/ srollins/courses/cs112-f08/web/notes/linkedlists.html. [Accessed
2016 May 15].
[39] R. J. Sutcliffe, Modula-2: Abstractions for Data and Programming Structures, Arjay Enter-
prises, 2005.
[40] W. Ulrich, "Application Package Software: The Promise Vs. Reality," Business Technology
and Digital Transformation Strategies , 31 August 2006.
[41] Nordic Semiconductor, "S110 nRF51 Bluetooth® low energy Peripheral SoftDevice," Nordic
Semiconductor ASA, Trondheim, 2014.
[42] R. Perlman, "network design folklore," Interconnections: Bridges, Routers, Switches, and
Internetworking Protocols, vol. 18, 1999.
[43] Comer , "They (protocols) are to communication what programming languages are to
computation," The Need For Multiple Protocols, vol. XI, no. 2, p. 177, 2000.
[44] Y. Fan, C. lvin, L. Songwu and Z. Lixia, "A Scalable Solution to Minimum Cost Forwarding
in Large Sensor," UCLA Computer Science Department, Los Angeles, 2001.
[45] Y. Jennifer, M. Biswanath and G. Dipak, "Wireless sensor network survey," Elvesier:Computer
Networks, no. 52, p. 2292–2330, 2008.

A Light-Weight Operating System For IoT Devices

Uploaded by

Copyright:

Available Formats

A Light-Weight Operating System For IoT Devices

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Light-Weight Operating System For IoT Devices

Uploaded by

Copyright:

Available Formats

A Light-Weight Operating System for

Internet of Things Devices

Embedded Computing Systems

Norwegian University of Science and Technology

PROJECT / MASTER THESIS

Department of of Engineering Cybernetics

Norwegian University of Science and Technology

Supervisor: Associate Professor Amund Skavhaug

Trondheim, June 2016

Summary and Conclusions

IOT Internet of Things

RAM Random Access Memory

CORBA Common Object Request Broker Architecture

SOA Service Oriented Architecture

OSI Open System Interconnection

API Application Porgramming Interface

HAL Hardware Abstraction Layer

CPU Central Processing Unit

TCP IP Transmission Control Protocol-Internet Protocol

SOC System on Chip

RTOS Real-Time Operating System

ROM Read-Only Memory

GUI Graphical User Interface

IPC Inter-process Communication

VLF Very Low Frequency

EHF Extremely High Frequency

ASK Amplitude Shift Keying

PSK Phase Shift Keying

FSK Frequency Shift Keying

DSSS Direct Sequence Spread Spectrum

FHSS Frequency Hopping Spread Spectrum

ISM Industrial Scientific Medical

GPIO General Purpose Input-Output

PCB Printed Circuit Board

RSSI Received Signal Strength Indication

AES Advanced Encryption Standard

CRC Cyclic Redundancy Check

DMA Direct Memory Access

I2C Inter-Integrated Circuit

SPI Serial Peripheral Interface

PSP Process Stack Pointer

MSP Main Stack Pointer

ISR Interrupt Service Routine

SVC Service Call

RISC Reduced Instruction Set Computing

MIPS Microprocessor without Interlocked Pipeline Stages

1.1 Prediction about IOT devices [5]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 A general structure of Embedded Systems [9]. . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Kernel Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.1 Structure of Hardware Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.1 Multitasking and Context Switching[37] . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.1 Structure of Linked List [38] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

7.1 Structure of Soft-Device[41] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

8.1 Classical Cluster Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

9.1 NRF518222 Development Kit and NRF51822 Dongle . . . . . . . . . . . . . . . . . . 100

2.1 Evaluation of Embedded Operating Systems . . . . . . . . . . . . . . . . . . . . . . . 33

8.1 Pros and Cons of Suggested Strcutres. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

9.1 Size of the Embedded Operating System . . . . . . . . . . . . . . . . . . . . . . . . . 102

4.1 Hal structure for GPIO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52