Performance Evaluation of Wi-Fi Networks

Nima Eivazzadeh Kaljahi

Specialization

Cloud-based Services and Operations

Thesis submitted for the degree of

Master in Applied Computer and Information
Technology (ACIT)
30 credits

May 2021
ii
Performance Evaluation of Wi-Fi Networks

Nima Eivazzadeh Kaljahi

© 2021 Nima Eivazzadeh Kaljahi

Performance Evaluation of Wi-Fi Networks

http://www.oslomet.no/

Printed: Oslo Metropolitan University

ii
iii
Abstract

iv
Abstract

Given the number of wireless devices in use is growing drastically, the bands are going increas-
ingly congested. Asking home users in that developed world about their wireless connection,
they are likely arguing that it seems to be getting worse not better. Access Points chooses a
radio channel randomly and therefore causes interfere the other wireless networks, while the
wireless traffic grows. The objective of this thesis is to examine what happens when two net-
works disturb each other and discuss this phenomena. Performance measurements of wireless
networks will be carried out. Throughput testing when two wireless networks working at the
same channel at the same time and interfering each other will be measured. In addition, the
measurement of single wireless network will be evaluated to produce different result in a con-
dition without the presence of interfering network to compare results.

v
Acknowledgment

vi
Acknowledgements

I wish to thank all the people whose assistance was a milestone in the completion of my Master
education and this thesis project:

• I wish to express my sincere gratitude to my supervisor, Madeleine Rønning, who has

the substance of a genius: she convincingly guided and encouraged me to be a profes-
sional and do the right thing even when the road got tough.

• My appreciation to Torleiv Maseng my other supervisor for all his effort and support.
For his helps, insightful comments and suggestions.

• My deep gratitude to my internal supervisor, Hårek Haugerud, for his kindness support
and encouragement during the difficult conditions in this thesis.

• My very special thanks to Kyrre Begnum as a brilliant and wonderful teacher of mine.
During these two years of study he always supported me in all situation and difficult
conditions.

• I would like to extend my sincere thanks to the Oslo Metropolitan University for offer-
ing me my desired Master education and all the facilities they provided throughout the
journey of my education.

• I would like to offer my special thanks to the University of Oslo for offering me the edu-
cation to learn more about Norwegian history, culture, society and language.

• I would like to express my sincere gratitude to my lovely family, my father and my

mother whom supported me throughout this journey everyday, every time.

vii
Acknowledgment

viii
Contents

List of Figures xi

List of Tables xiii

1 Introduction 3
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 The problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Background and Research Work 9

2.1 Factors affecting wireless network performance . . . . . . . . . . . . . . . . . . . 9
2.2 Fragmentation and Aggregation: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Carrier Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 IEEE 802.11 MAC Frame: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 The MAC sublayer : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5.1 Distributed Coordination Function . . . . . . . . . . . . . . . . . . . . . . . 17
2.5.2 Backoff algorithm BEB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.6 RTS/CTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6.1 RTS/CTS handshake in action . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6.2 Exposed Terminals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.7 Experimental platform developed by other researchers . . . . . . . . . . . . . . . 22

3 Approach 25
3.1 Experiments design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.1 Testbed design and hardware used for the measurement . . . . . . . . . . 26
3.2 Throughput testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 The mechanisms of the throughput testing . . . . . . . . . . . . . . . . . . 28
3.2.2 Theory of operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Tools used for capturing data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Wireshark Filtering feature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Operating system of Raspberry Pie . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5.1 Rasspbery Pi’s MAC address . . . . . . . . . . . . . . . . . . . . . . . . . . 32

ix
Acknowledgment

4 Results and Analysis 33

4.1 Range limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Result of experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

5 Discussion and Future Work 47

5.1 Problems and challenges occured during measurement . . . . . . . . . . . . . . . 47
5.1.1 Equipment problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.1.2 The measurement tool used before iperf3 . . . . . . . . . . . . . . . . . . . 51
5.2 Result of measurement accumulated by other researcher . . . . . . . . . . . . . . 52
5.3 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.4 Challenges occured during special days . . . . . . . . . . . . . . . . . . . . . . . . 53

6 Conclusion 55

7 Appendix 57
7.1 The Python script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7.2 Wireshark filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Bibliography 61

x CONTENTS
List of Figures

1.1 Collision Avoidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2 Hidden nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 CSMA/CA with RTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1 Interference Pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2 Flow diagram of slotted CSMA/CA . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 IEEE 802.11 MAC Frame structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 MAC sub-layer and the related physical layer standards for IEEE 802.11 . . . . . 16
2.5 Event sequence in DCF CSMA/CA . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 Basic access mode of DCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.7 Principle of RTS/CTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.8 Schematic of BEB algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.9 Hidden Node . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.10 RTS/CTS handshaking with ACK . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.11 Exposed Terminal Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.12 Measurement setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.1 Single network, One and Five meter distances between client and access point. . 29
3.2 Controlled and uncontrolled interfering networks are available and they are
working in the same environment. One and Five meter distances d (distance)
between two wireless networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.1 Throughput without the presence of interfere network in two different distances. 38
4.2 The percentage of frames re-transmitted without the presence of interfere
network in two different distances. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Frame re-transmitted without the presence of interfere network in two different
distances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.1 Single network with USB wireless card on a client . . . . . . . . . . . . . . . . . . 48

5.2 Single network and monitor mode device captures traffic information from air. . 49
5.3 Steps to capture air traffic with the presence of Client2 ans AP2 as a interfere
network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

xi
Acknowledgment

5.4 Steps to capture air traffic without the presence of Client2 and AP2 as a interfere
network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

xii LIST OF FIGURES

List of Tables

3.1 The measurement architecture with one network . . . . . . . . . . . . . . . . . . . 29

3.2 The measurement architecture with two networks . . . . . . . . . . . . . . . . . . 31

4.1 Result of experiment when two wireless networks have One meter distance
from each other. Channel 108 pkt size 1470. Frequency 5540MHz (Band 5GHz).
RTS threshold 2346 bytes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 Result of experiment when two wireless networks have Five meter distance
from each other. Channel 108 pkt size 1470. Frequency 5540MHz (Band 5GHz).
RTS threshold 2346 bytes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Result of experiment when two wireless networks have One meter distance
from each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz).
RTS is Off. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 Result of experiment when two wireless networks have One meter distance
from each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz).
RTS is Off. Repeated measurement of table 4.3. . . . . . . . . . . . . . . . . . . . . 41
4.5 Result of experiment when two wireless networks have Five meter distance
from each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz).
RTS is Off. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.6 Result of experiment when two wireless networks have One meter distance
from each other. Channel 108 - pkt size 6560. Frequency 5540MHz (Band 5GHz).
RTS is Off. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.7 Result of experiment when two wireless networks have Five meter distance
from each other. Channel 108. Pkt size 6560. Frequency 5540MHz (Band 5GHz).
RTS is Off. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.8 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 2100.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.9 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 2100.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

xiii
 Acknowledgment

4.10 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 1470.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.11 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 1470.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.12 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 140. Pkt size 1470.
Frequency 5700MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.13 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 140. Pkt size 1470.
Frequency 5700MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.14 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 6560.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.15 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 6560.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.16 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 1470.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.17 Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 1470.
Frequency 5540MHz (Band 5GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

LIST OF TABLES 1
Introduction

2 LIST OF TABLES
Chapter 1

Introduction

Today, wireless technologies has increased speed, enhanced capabilities and expanded to de-
liver peer-to-peer connectivity with high speed performance, smart and secure home network
services[1]. Wireless communication has opened vast new avenues for entertainment. Devices
such as smartphones are equipped with applications for downloading and reading books,
newspapers, streaming games, movies, television and live sporting events. People with smart-
phones now have a way keep themselves almost endlessly entertained while on the go or to
fill the time between appointments and classes.

Wi-Fi ® 6 is the latest generation of wireless network technologies offering great capacity, ef-
ficiency, and performance for advanced connectivity due to [2]. According to Wi-Fi Alliance,
Wi-Fi 6 will observe strong global adoption across PCs, access points, smartphones as well as
IoT devices in enterprise, homes and public areas with almost nearly 2 billion Wi-Fi 6 devices
shipments in 2021 .

In wireless technology wireless devices communicate by radio frequency band. Frequency

bands in 2.4, 5 as well as newly emerged 6 GHz, (Wi-Fi 6E) 1 are available for wireless commu-
nication at the time of writing the report. The 802.11 standard defines fourteen channels for
use by 2.4 GHz. Each channel is 20 MHz wide with an additional 1 MHz on the low and high
end of the channel for inter-channel spacing. This result in each channel requiring 22 MHz
of bandwidth according to [3]. Channels are spaced 5 MHz apart. Since of this spacing it is
not possible to actually use 11 channels in the same location since the 22 MHz width would
overlap and cause interference [3].

In the 5GHz band, channels are ranging from 36 up to 165 MHz wide [4]. Same as the 2.4
GHz wireless channels, each 5GHz channel is 5 MHz wide. There are 23 channels specified for
802.11 wireless network, unlike the 2.4 GHz channel definitions, only non-overlapping chan-
nel numbers are used for device configuration according to [3].

1 https : //www.wi − f i.org/countries − enabling − wi − f i − 6e

3
Introduction

A question arises as to how interference problem occurs generally? Interference can cause
from our own network or neighbor’s wireless devices if they are working in the same channel,
microwaves or may a radar systems. Signal to noise ratio is one factor affects the interference
problem in wireless networks due to [5]. In general, SNR compares the level of a desired signal
to the level of background noise. SNR is defined as the ratio of signal power to the noise power
measured in decibels according to [6]. For instance, a client can be right next to an access point
with excellent signal, but unable to keep a connection if the signal from another wireless net-
work or any other type of radio-frequency device is too high, meaning that signals from other
devices are just noise to your device.

By the given information in mind the importance of wireless network is more serious in the
society so therefore it plays an important role in people’s daily activities.

As more wireless devices are connected and wireless access points are densely deployed in
the scarce frequency spectrum, the failure probability of packet transmission is expected to in-
crease due to interference from other devices. This is because the the 2.4 Ghz band is already
congested and the 5 Ghz band will be congested soon, the wireless environment might suffer
severe interference from unintentional wireless devices.

The increasing deployment of access points for wireless services result in more interference
from neighboring access points, thus noticeably degrading the wireless network performance
and damaging the network security as well as quality of service.

1.1 Motivation

In home environment users use various type of devices such as portable computers, mobile
phones, tablet computers, intelligent televisions, intelligent fridges and IoT systems which all
requires to have Internet connectivity via home-based wireless network settings. By this as-
sumption in mind, the quality of service in home-based wireless networks is a significant factor
for home users who prefer to use wireless network technologies at home.

Access to various type of devices through a cable connection would not be efficient and pos-
sible at home environment. Therefore, wireless technologies can bridge the gap through its
outstanding services allow users to connect themselves to those devices without any computer
network cables. This feature is a positive point of wireless network technologies so therefore,
it plays a positive role in people’s daily life as it makes connectivity simple and more conve-
nient for them. The motivation behind this project is to analyze wireless network performance
specifically in home environment with the presence of controlled and not controlled interfer-

4 CHAPTER 1. INTRODUCTION
 Introduction

ing wireless networks to observe the behaviour of the network. More specifically to acquire the
better understanding of the factors in place which affect the performance of wireless network.

Home users use wireless routers to connect their own devices to the Internet. There are differ-
ent type of access points working in the apartment buildings almost near to each other with
different distances and different radio frequencies and channels.

In general, there are some parameters affect the performance of the wireless networks. For
example, transmit power, received signal from access point (signal strength), distance between
access point and clients, busy channel, external heavy noise comes from neighboring wireless
devices and a very close distance between a pair of wireless networks.

In wired network carrier-sense multiple access with collision detection is a media access con-
trol method was used in early Ethernet technology for local area network. It uses carrier-
sensing to defer transmissions until no other stations are transmitting [7]. In wireless network
this method is carrier-sensing multiple access with collision avoidance.

In wireless network, nodes have random access to a channel. This means they can transmit
whenever they need to, but they can interfere with other nodes transmitting data, causing
data loss. CSMA listens to the shared channel, in case the medium is not idle (someone else
transmitting), it waits for transmission to become complete. Then, start transmitting data to
avoid collision. It is possible to mention that collisions can still occur if two nodes transmit at
the same time. Collision Avoidance waits for a random amount of time when the channel is
busy to avoid collision. Figure 1.1 illustrates the process.

Figure 1.1: Collision Avoidance

CHAPTER 1. INTRODUCTION 5
Introduction

In wireless network, sender expects an acknowledgment message (ACK) back from the receiv-
ing device to confirm it arrived correctly. Therefore, if it has not arrived in a set amount of
time, the data is re-transmitted.

A problem of CSMA for wireless networks is the hidden nodes problem. For instance, as fig-
ure 1.2 illustrates node B and node C can sense each other as they are in the same range of
each other. In addition, they can communicate with the router since they are all in the same
range. On the other hand, Node A is communicating with access point at the same time as
well, but B and C cannot see that node. This is because Node A is far enough and it is not on
the same range of B and C. This adaptation requires a node to send a Request to Send message
and receive back a Clear to Send message from access point before transmitting. Therefore, to
overcome the hidden node problem, RTS/CTS mechanism is implemented at access point in
conjunction with the CSMA/CA scheme according to [8]. With RTS/CTS mechanism packet
can send with less risk of collision, but it adds overhead to each packet, and will worsen con-
gestion. Therefore, it might be deactivated for small packets. Figure 1.3 illustrates process of
CSMA/CA with RTS mechanism.

Figure 1.2: Hidden nodes

6 CHAPTER 1. INTRODUCTION
 Introduction

Figure 1.3: CSMA/CA with RTS

[8]

Since this research study tries to evaluate performance of home-based wireless networks as well
as measuring interfering problems in home environment. Therefore, it tries to use real wireless
network devices instead of using simulated software such as NS3 or Mininet to observe the real
behaviour of the wireless network performance. Thus, as these simulation tools simplify the
wireless medium and the result obtained from these environments is not correctly translatable
to real-life environment problems, Raspberry Pie devices is used to make a real wireless
network with real devices rather than simulated environment to observe the performance in
real environment.

1.2 The problem statement

In IEEE 802.11 the channel access is controlled by a carrier sense procedure. This includes the
clear channel assessment (CCA) which is a function with CSMA/CA and performs the channel
access by observing the channel for ongoing transmissions. It prevents the node form sending
if transmissions of other networks are sensed on the same channel. The following two cases
result in reducing rates due to [9]:

• Any energy above a certain threshold triggers the channel as busy and the channel will
wait. This causes the rate to be reduced.

• On the other hand, if no signal is detected, transmission starts. It is possible that it should
not have started and failed to detect that the channel was busy. This would result in many
faults in the frames. This will result in re-transmission of frames.

The objective of this thesis is to examine what happens when two networks disturb each other
and discuss this phenomena.

Performance measurement with the presence of single wireless network as well as two pair of
wireless networks will be conducted through a throughput testing application. For instance,

CHAPTER 1. INTRODUCTION 7
Introduction

when one single pair of wireless network works stand-alone in two certain distances. More-
over, when two pair of wireless networks work beside each other horizontally on the same
channel with the same configuration at the same environment. The purpose of measurement
is to observe the behaviour of the network, variations and numbers.

The approach chapter will explain clearly the project’s testbed architectures through different
type of visual illustrations and figures. Subsequently, the result of measurements relevant for
each architecture introduced in the approach chapter will be reported in result chapter.

8 CHAPTER 1. INTRODUCTION
Chapter 2

Background and Research Work

In order to understand the interference problems in home-based wireless networks, it is crucial

to have adequate understanding of the problem domain. This chapter explain interference
problems, medium access, career sensing as well as RTS/CTS mechanism obtained from
scientific papers as well as online resources. Moreover, it would be a comparison between the
work has been developed and implemented in this project and other research work defined
by other researchers regarding the problem domain. Moreover, this chapter tries to present
knowledge in data link layer and connect them to the work accumulated in the approach
chapter by reviewing scientific paper.

2.1 Factors affecting wireless network performance

Wireless interference occurs when something disrupts or weakens the signal coming from an
access point. The most typical channel for wireless connection is in the 2.4 GHz band. This
channel is shared with other devices in home environment, the limited channel in 2.4 GHz
increases the interference further. Interference from other radio signals can affect network
performance. The higher interference between neighbors the more congestion and low speed
throughput occurs in a wireless networks.

Due to [10] there are some factors affects the performance of wireless networks. Below are a
few example of them .

• Antennas:

Wireless network performance can be affected by the antennas in an environment service

equipment. Antennas has a significant impact on network coverage. For instance, if
antennas are near some sort of structure, such as a metal grid or cement beam, this can
affect wireless transmissions and subsequently reduce the performance.

• Power Level:

9
Background

Power level of wireless equipment also influence speed. Power level is transmitted by an
access point is considered as wireless coverage. The transmit power of an access point
radio is proportional to its effective range. The higher the transmit power, the farther a
signal can travel, and the more obstructions it can effectively penetrate. Stefan [11] ar-
gues that a common approach to assess the quality of the network is the received signal
strength indicator (RSSI). The RSSI corresponds to the signal strength at the receiving
antenna. Duo to [12], it has been widely realized that the link reliability has significantly
related to received signal strength indicator known as RSSI (Received Signal strength In-
dicator) or signal-to-interference-plus-noise ratio known as SINR. In addition, external
interference makes it unpredictable which is different from the previous understanding
that there is no tight relationship between the link reliability and RSSI or SINR.

To reduce interference, maximum transmit power is reduced. In addition, observing that

nodes should not adapt their rates due to losses during congestion (channel busy time)(the
fraction of time the medium is utilized) can be used as metric. Noting that in this case,
the evaluation of channel busy time has dependency on the underlying assumption that
physical carrier sense mechanisms, which play an important role in avoiding packet col-
lision [13].

Woo [12] argues that, both the link distance and Signal-to-Interference Ratio (SINR) are not
strongly correlated with frame loss rates. Woo et al. They also claim that received signal
strength indicator (RSSI) is a good indicator to predict link performance in (rural) mesh
network that barely have external interference.

• Interference:

Woo et al [12], in wireless networks, links usually do not permanently stay at the state
of link reliability 1.0 or at the opposite state of link reliability 0.0 because of external in-
terference, multi-path fading, where the link reliability is explained in terms of frame
delivery rate (FDR). Also , each wireless link rate additionally turns out to have an in-
termediate range of link reliability in between 0.0 and 1.0. This means that links do not
always successfully transmit data or do not successfully drop data in a given interval.

Hwangnam et al [12]. Understanding of irregular variation of wireless links enables net-

work protocols or systems to use optimal link or different classes of wireless link accord-
ing to the current link state or required link quality. Moreover, Woo et al [12] claims that,
link performance can be predicted by RSSI which can be easily translated to SINR with a

10 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

constant noise floor in rural mesh networks without having a external interference.

Due to [14] powerful signal can interfere with neighboring devices even if they are
on different channels (frequencies). Access points can interfere with each other, even
if there is enough space distance between them. Interference can also emanate from
adjacent channels. Thus, in this case the design of an 802.11 system’s, RF sub-system
and digital filtering can greatly affect the performance of the wireless network. Also,
the physical design of a wireless network can overcome many of the consequences of
in-band interference. The performance of wireless network is determined by the signal -
to interference ration (S/I or SIR), which is defined as the ratio of the data signal to the
interference signal. The signal-to-interference ratio (SIR) measures the aireless signal
in comparison with how much co-channel 1 interference is present from other radio
transmitters due to [15]. SIR is usually more critical to WLAN performance than the
signal-to-noise (SNR). Figure 2.1 illustrates the mentioned concept [16].

[16]

Figure 2.1: Interference Pattern

According to [17] IEEE 802.11 uses many mechanisms to mitigate noise and interference,
thus it is natural to ask whether 802.11 links are already as robust to interference as can
reasonably be expected. Those mechanisms are as follows:

1. A MAC protocol that avoids collision.

2. lower transmission rates that accommodate lower. signal-to-interference-plus-noise

ratios (SINR).

3. signal spreading that tolerates narrow-band fading and interference.

4. PHY layer coding for error correction.

5. RTS/CTS mechanism to reduce the risk of collision in wireless network.

1 co-channel interference or CCI is cross-talk from two different radio transmitters using the same channel.

CHAPTER 2. BACKGROUND AND RESEARCH WORK 11

Background

2.2 Fragmentation and Aggregation:

Fragmentation aims to improve wireless performance in a cluttered environment due to

[18]. Fragmentation break up packets into smaller pieces, for a higher chance of successful
transmission. A receiver has to put all these fragments back together to form the original
packet. Moreover, aggregation is used to acknowledge blocks or groups of packets, which
cuts down the requirement to acknowledge every single packet, enhancing efficiency and
performance. Due to [19] frame aggregation is a function that combines several frames
into a single large frame for transmission. Frame aggregation has many benefits: first,
transmitting large frame leads to higher throughput than transmitting small frames. Second
and most important benefit is the reduction of timing and header overheads that are required
to transmit a frame by the MAC distribution coordination function. Therefore, by using frame
aggregation, these overheads are squeezed and only few overheads have been used to transmit
the aggregated frame. Aggregation can be performed either at the packet level or at the
frame level. It is called packet aggregation of it is performed at higher layers such as IP and
application layers. However, it is called frame aggregation if it is performed at the lower layers
such as PHY and MAC layer. The frame level aggregation provides more control over the
transmitted frames and exhibits an efficient partial re-transmission. Moreover, aggregation
at the MAC level is the widely used aggregation where the MAC headers overhead can be
squeezed or even removed. Also, the channel access can be optimized by reducing the timing
overhead such as backoff and message exchange overhead such as ACKs.

2.3 Carrier Sensing

According to [20] carrier sense is the fundamental part of most wireless networking stacks in
wireless local area and sensor networks. As growing number of users and more demanding
applications uses wireless networks to their capacity limits, the efficacy of the carrier sense
mechanism is a key factor in determining wireless network capacity.

A question arises as to what is the basic idea of carrier sensing? due to [20] in career sensing
before transmitting a packet a sender listens to the channel and evaluates whether a nearby
node is transmitting a packet or not. Therefore, if no nearby node is transmitting a packet the
sender transmits a packet immediately. If a nearby node is transmitting, the sender defers and
it waits for a few time after the end of the intervening transmission.

Carrier sense is a part of the medium access control MAC layer of the radio stack. Well-
informed MAC decisions are crucial for maximizing the capacity of a broadcast radio medium.
Failed transmissions not only waste energy, but also have potential to corrupt other transmis-
sions in the network, reducing aggregate capacity. Deferring a transmission has the potential
of wasting a good transmission opportunity, therefore reducing capacity. Due to [20] in carrier

12 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

sensing the sender and receiver are in different locations, and the sender makes the carrier
sense decision based on local information .

Contention Window (CW) is related to idle channel availability by using Clear Channel Assess-
ment (CCA). Initial value of CW is 2, that is a node needs to pass the channel availability for
two successive times before declaring the channel idle according to [21]. Figure 2.2 illustrates
the detailed flow diagram of CSMA/CA.

[21]

Figure 2.2: Flow diagram of slotted CSMA/CA

Jamieson et al[20]. Although carrier sense improves link qualities at all traffic loads, but it
leaves room for performance improvements. In addition, they claim that carrier sense can ac-
tually reduce capacity under extreme loads. Jamieson claim that carrier sense is not always a
good predictor of transmission success since it relies on channel measurements at the sender to
infer the probability of reception at the receiver. However, in many cases, no correlation exists
between channel conditions at the sender and at the receiver. Thus, this lack of correlation
is often because of exposed terminals, the aggregate effect of distant nodes raising the noise
floor. In addition, they claim that local level carrier sense may perform poorly when exposed
terminals are present.

The IEEE802.11 specify two mechanisms for sensing whether a medium is busy or not, those
are as follows:

• Virtual carrier sensing

• Physical carrier sensing

CHAPTER 2. BACKGROUND AND RESEARCH WORK 13

Background

In virtual carrier sensing the medium is defined busy by reading the Network Allocation
Vector (known as NAV) field present in the MAC frame. This filed can be read when using
RTS/CTS control frames.

The physical CS (PHY) is a different mechanism. In IEEE 802.11 networks, physical carrier
sensing is defined by the clear channel assessment (CCA) function monitor the channel to de-
fine whether the channel is free or is in use.

Regarding RTS mechanism Jamieson et al [20]. Represented that imagine two transmitters, for
instance node A and B, are both within radio range of each other. The intended recipients of
their transmissions, nodes A’ and B’ respectively are within range of only one transmitter, and
thereby could simultaneously receive a packet from the intended transmitter. In this environ-
ment carrier sense would only allow one transmission to take place. Whichever node lost the
CSMA contention period would sense a busy channel and wait for the other nodes’s transmis-
sion to complete. In addition , carrier sense may be a poor predictor of transmission success
if interference comes from large number of distant nodes rather than a few local neighbors
nodes. He claims that when interference is local and nodes are within each other’s transmis-
sion range, carrier sense or RTS/CTS exchange may be a good method of contending for the
channel.

There are three different methods that can be considered to see whether a channel is busy or
not.

1. Energy detection.

2. High rate PHY signal.

3. Direct sequence spread spectrum (DSSS) signal.

DSSS signal looks at an actual DSSS signal and report the medium as busy if a signal is detected
regardless of the received energy. DSSS signal is widely used both in military and commercial
communication. This is because of the low power spectrum density, it is very difficult to detect
in non-cooperative communication. It has very low power signal spectral density, which result
in the transmission below the noise level [22].

Huehn et al [13]. In CSMA wireless networks several mechanism are exists, mechanisms such
as transmit rate, transmit power control. According to Huehn carrier sense aim is to max-
imize throughput performance, while these presented mechanisms operate independently,
they present high dependency affects the optimum of transmission decision . For instance,
A packet can be transmitted at a high rate in case the SNR (Signal to Noise Ratio) at receiver is

14 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

high. Otherwise, a lower transmit rate achieves more robust communication.

2.4 IEEE 802.11 MAC Frame:

The MAC layer frame consist of nine fields due to [23]. Figure 2.3 illustrates the basic structure
of an IEEE 802.11 MAC data frame along with the content of the frame control filed.

[23]

Figure 2.3: IEEE 802.11 MAC Frame structure

Type is a two bits long filed which determines the function of frame, i.e management (00), con-
trol(01) or data(10), the value 11 is reserved due to [23].

Sub-type is a four bits long filed which indicates sub-type of the frame. For example, 0000 for
association request, 1000 for beacon.

To DS is a one bit long field indicates that destination frame is for DS (distribution system).

From DS is a one bit long filed indicates frame coming from destination.

Retry bit is one bit long field. It represents if the current frame is a re-transmission of an earlier
frame, then this filed is set to 1. This means that when client transmit a frame to its access point,
but didn’t received acknowledge message from its access point, then client assumes frame was
lost in between of transmission so therefore re-transmits frame again. This research study
will look into Retry field through the measurements to observe the probability and number
of frames re-transmitted from client to its access point in different measurements. This is to
observe if channel is busy or if client received weak signal or signal is lost, then how many
error and re-transmission of frames occurs. For instance, how many Retry bit are exist when a
Pair of wireless network works stand-alone and how many are exist when two Pairs of wireless
network work beside each other and result in interference. Approach chapter will explain the
matter more in detail.

CHAPTER 2. BACKGROUND AND RESEARCH WORK 15

Background

2.5 The MAC sublayer :

IEEE 802.11 defines two MAC sub-layers due to [24]:

• The distribution coordination function (DCF).

• Point coordination function (PCF).

Figure 2.4 illustrates MAC sub layers and physical layer standards for IEEE 802.11.

[24]

Figure 2.4: MAC sub-layer and the related physical layer standards for IEEE 802.11

16 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

2.5.1 Distributed Coordination Function

In IEEE 802.11 family of standards DCF protocol, controls access to the physical medium. A
station must sense the status of the wireless medium before transmitting.

According to [25] The DCF uses CSMA/collision avoidance mechanism to control access to
the shared wireless medium. This is because collisions are difficult to detect in wireless en-
vironment, therefore, a backoff-based collision avoidance technique, rather than the collision
detection technique which is common in Ethernet standard is used. Each wireless station/user
first listens to the wireless medium to detect transmissions. If medium is sensed to busy, the
station waits until the ongoing transmission is over. If the medium is detected to be idle for a
distributed inter-frame space (DIFS) interval, user enters a backoff procedure. In the backoff
procedure, user selects a random backoff time (in slots) from a contention window, and starts
decreamenting a backoff counter for each slot that is sensed to be idle, while counting down,
another user begins transmitting. User in backoff mode suspends its counting, until the trans-
mitting user finishes and the medium is sensed to be idle for a DIFS duration, and resumes
its countdown thereafter. Once the backoff interval expires, the user begins transmission. The
value of the random backoff interval is chosen from an interval called the contention window




CW which lies between two pre-configured values, CWmin and CWmax . The contention
window is set to ( CWmin ) at the first transmission attempt, and doubles after each unsuccessful at-
tempt, until it reaches ( CWmax ). The contention window is reset to ( CWmin ) after every successful
transmission. This procedure illustrated in figure 2.5 as well as figure 2.6.

[25]

Figure 2.5: Event sequence in DCF CSMA/CA

[26]

Figure 2.6: Basic access mode of DCF

In DCF configuration, a contention window is set after a frame is transmitted. This is considered to
avoid collision. The window defines the contention time of various stations who contend with each
other for access to channel. Thus, each of the stations cannot garb the channel immediately, rather the

CHAPTER 2. BACKGROUND AND RESEARCH WORK 17

Background

MAC protocol uses a randomly chosen time period for each station after that channel has undergone
transmission [27]. For CSMA/CA medium access, a backoff mechanism with a contention window
comparable to IEEE 802.11 is used in order to reduce the collision probability. As collisions cannot be
avoided completely, the receiver of a frame sends an acknowledgment in order to enable the transmitter
to detect transmission errors.

Cheng et al [26] argue that CSMA/CA in IEEE 802.11 provides confirmation frame ACK to ensure
the check of frame lost and re-sent. For further avoiding conflict, RTS/CTS + ACK four handshakes
protocol is imported. RTS/CTS protocol is a common mechanism to reserve channel by the way of
handshakes between sender node and receiver node. Figure 2.7 shows the process of data transmission
between the sender and the receiver by use of RTS/CTS + ACK + DATA, called handshakes protocol
and the situation of other nodes setting NAV. Other nodes renew the value of NAV by use of the filed
duration of RTS/CTS frame after receiving this frame. The node could not send frame before the value
of NAV equal zero. When only the value of NAV equal zero, the node could send data frame by the
way of DCF.

[26]

Figure 2.7: Principle of RTS/CTS

2.5.2 Backoff algorithm BEB

Due to [26] In IEEE802.11 WLAN, all nodes can carry out carrier sense. Each node operates sending
action according to backoff counter when data packets are trying to send at the first time. The con-



tention window CW is set to the minimum CWmin before sending, and transmission is carrying


out in equal probability at a choice time between 0 , CWmin . Therefore, when a channel is free for an
interval, the backoff counter would decrease 1. So, when the channel is sensed busy, the backoff couner
would be freeze until the channel is sensed free for DIFS time, and then the backoff counter would be
unfreeze and continue to sense the channel. According to figure 2.8 node A and node B are share the
channel.

18 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

[26]

Figure 2.8: Schematic of BEB algorithm

In BEB algorithm, backoff window CW of a node increase redouble until maximum CWmax when



conflict occurred, backoff window CW decrease until minimum CWmin when sending is success-


full. Below formula (2.1) explains the renew regular of CW in BEB algorithm.

(

Finc = Min(2 ∗ CW, CWmax )


(2.1)

Fdec = CW



CW is the value of backoff window, CWmin and CWmax are set according to the channel load.



According to [26] typically CWmax = 1024 , and CWmin = 2.

Taijun li et al [26] claims that the size of backoff window depends on the times of experienced conflicts.
Contention window increases drastically with the increasing of re sending times, reducing the conflict
probability. However, this leads to the node which is the last one sending successful has the smallest
backoff time, while is at a disadvantage in next contention, and the probability of the node sending
failure access channel again is significant decreasing. Therefore, BEB algorithm always gives the node
which is the last one sending successfull the greatest priority.

2.6 RTS/CTS
Due to [28] Request to send / clear to send is the optional mechanism used by the 802.11 wireless net-
working protocol to reduce frame collision introduced by the hidden node problem. Hidden nodes are
the nodes that are not in the range of other nodes or a group of nodes. Each node is within communica-
tion range of the access point, but the nodes cannot communicate with each other as they do not have
physical connection to each other. For instance, in a wireless network, it is possible that the node at the
far edge of the access points’ range, known as r, can see the access point, but it is unlikely that the same
node can see a node on the opposite end of the access point’s range, r2. Thus, these nodes are known as
hidden

Figure 2.9 illustrates the hidden nodes problems.

Kanapi et al [28]. The problem is when nodes r and r2 start to send packets simultaneously to the
access point. Since node r and r2 cannot sense the carrier, Carrier Sense Multiple Access with Collision
Avoidance does not work. To solve this problem, handshaking is implemented in conjunction with the
CSMA/CA scheme.

CHAPTER 2. BACKGROUND AND RESEARCH WORK 19

Background

[28]

Figure 2.9: Hidden Node

2.6.1 RTS/CTS handshake in action

• A is the source which is in the range of B, D and C.

• B is the destination which is in the range of A, D and E.

• A is the source which is in the range of B, D and C.

• B is the destination which is in the range of A,D and E.

• B sends ACK after receiving one data packet.

• Improves link reliability using ACK show in the below figure.

[28]

Figure 2.10: RTS/CTS handshaking with ACK

IEEE 802.11 uses RTS/CTS acknowledgment and handshake packets partly overcome the hidden node
problem. RTS/CTS is not a complete solution and may reduce throughput even further, adaptive ac-
knowledgment from the base station can help too.

Due to [29] the sender first sends the RTS frame to reserve the channel before its transmission, and
upon receiving the CTS frame from the receiver, the normal packet transmission and the ACK response
proceeds.

2.6.2 Exposed Terminals

Kanapi et al [28]. In wireless networks, the exposed node problem occurs when a node is prevented
from sending pakcets to other nodes due to a neighbouring transmitter. For instance, consider 4 nodes
labaled R1, S1, S2 and R2, where the two receivers are out of range of each other, the two transmitters
in the middle are in range of each other as it presents in figure 2.11. If a transmission between node
S1 and node R1 is taking place, node S2 is preventing from transmitting to node R2 as it concludes
after carrier sense that it will interfere with the transmission by its neighbor node S1. Due to [28] node

20 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

R2 could still receive the transmission from node S2 without interference since it is out of range from S1.

[28]

Figure 2.11: Exposed Terminal Problem

CHAPTER 2. BACKGROUND AND RESEARCH WORK 21

Background

2.7 Experimental platform developed by other researchers

Michael et al [9] prepared the experimental architecture with the presence of two pair of networks
which they call pair 1 and pair 2. In order to evaluate the influence of adjacent interference on 802.11
throughput they prepared two competing pairs. Figure 2.12 illustrates the scenario:

[9]

Figure 2.12: Measurement setup

AP and client in each pair are placed side by side to achieve the maximum SNR (Signal to Noise ration).
In addition, in order to generate traffic and maximize the throughput between pairs and flood the chan-
nel they used iperf throughput tester application. In addition, UDP packet with the default length 1470
bytes sent to the link to transmit traffic between client and its AP’s. In order to avoid the influence of
rate adaption schemes in all experiments the transmission rate is fixed. In addition, based on the fact
that APs and clients in typical residential deployments are not communicating in such a close proximity
Michael et al performed measurements for different distances d between the interfering pairs, while the
distance B between AP and client in each pair was increased to three meter. Michael argues that the
distance between competing access point and client pairs is crucial for the interference between them.
Therefore, they argue that a higher distance between adjacent WLAN links result in a higher signal
attenuation (weaker) and therefore to a better SINR. The same transmission power is used during the
measurement.

They argue that for different distances between the pairs it can be discovered that even very short links
are affected from adjacent channel. In addition Michael et al [9] claim that the interference generated by
one pair has only a minor influence on the other. They claim that in the real life it is very unlikely that
a client and an access point are placed directly side by side, so the inter-link distance B between access
point and client increased to 3 meter. Therefore, the measurement setup they implement is based on
assumption of general WLAN use case which often applied in residential or office environment. Also,
based on the result they accomplished they claim that the distance d has almost no influence on the
achievable link throughput. Therefore, for the purpose of clarity only the throughput of one link pair 1
is given.

According to Michael et al [9] results they stated that the usage of Partially overlapping channel (POC)
in 802.11g will not automatically lead to improvement in the cumulative throughput. Also, they cliam
that concerning the channel access scheme the adjacent channels have limited scope of application. So,

22 CHAPTER 2. BACKGROUND AND RESEARCH WORK

 Background

this behaviour makes the practical implementation of POC in 802.11 g pointless. They summarized and
concluded that the POC has an advantage in 802.11 b, but the restricted application of CSMA/CA and
the different PHY makes POC in 802.11 gnot recommendable any more.

CHAPTER 2. BACKGROUND AND RESEARCH WORK 23

Approach

24 CHAPTER 2. BACKGROUND AND RESEARCH WORK

Chapter 3

Approach

This chapter provides an overview of the techniques used in this research. It covers the considered
ways that tries to answer the problem statement followed by design of experiments. Moreover, the
different type of testbed architectures will be shown in this chapter in order to interpret the different
configurations setup for devices used for the measurements of this research study. In it worth noting
that the idea behind designing the testbed used to run the experiments is a consultation between the
author and supervisors as well as scientific paper such as [9].

Based on a discussion author had with supervisors regarding the studying and analyzing measurement
of performance of wireless networks, those agreed that to work on data link layer rather than network
layer for measurement and capture frames through Wireshark. Therefore, result of the measurements
produced in result chapter is a result measured by Wireshark.

In order to properly answer the problem statement a proper experimental design should be considered.

3.1 Experiments design

There are different strategies doing the measurement which those are as follows:

• First and second measurement consisting of one access point and one client. An access point and
client in first measurement has one meter distance from each other. In addition, in the second
measurement access point and client has five meter distance from each other. This measurement
has completed in a channel which observed is least busy in the test environment. This means
that there are a few not controlled access points working on that channel. In the time of testing
channel 108 in 5 GHz frequency band observed a least busy channel.

• Third and forth measurements consist of two wireless networks (networks such as Client1 and
AP1 as well as Client2 and AP2). Distance between wireless networks are one and five meters. In
the third measurement Client1 and AP1 network has static distance which is one meter between
client and access point. Also, the interfere network Client2 and AP2 has two certain distances such
as one and five meter from Client1 and AP1 network. The reason why two wireless networks has
two distances from each other is to observe results when Client2 and AP2 interfere Client1 and AP1
in different distances.

25
Approach

• Fifth and sixth measurements were completed by single Pair of network (Client1 and AP1) in one
and five meter distances between client and access point, but in a clear channel. clear channel is
a channel that there are not any access points working on that channel in the environment test
has been completed. WiFi Analyzer mobile application was used to define the the channel in the
environment.

• In seventh and eight measurement the interference Client2 and AP2 network works beside Client1
and AP1. This means that two wireless networks working beside each other in a clear channel.
Same as third measurement the interference wireless network which is Client2 and AP2 works in
one and five meter distance from Client1 and AP1 network.

3.1.1 Testbed design and hardware used for the measurement

In this project five Raspberry Pi devices are used. Two Raspberry Pi’s are working as access points and
two others are working as clients. In addition, one Raspberry Pi works in Monitor mode to capture the
MAC Frame fields in the data link layer.

The aim of the measurement is to capture the low level data packet which is called frames on the MAC
layer to observe MAC frame fields most specifically Retry bit. This work is possible when the state of
the wireless card changes from Manage mode to Monitor mode. The wireless card in Monitor mode can
capture every packets travels on the air, but wireless card in Manage mode can only capture packets
that have device’s MAC address in network layer. Therefore, it can capture specific type of packets
which is understandable by the MAC address of a device.

Through Monitor mode it is possible to observe the number of Retry bit in order to understand that
how many data packet has been transmitted more than once from client to access point. This means
that this type of frames transmitted once, but server could not interpret the packet. Therefore, client
assume that the packet lost in between and requires re-transmission to server again. For this purpose
the first experiment consists of Client1 and AP1 network with Raspberry Pi devices was completed and
Wireshark application captured those required information independently through the Monitor mode
wireless card.

It is worth nothing that the Raspberry Pi’s wireless card by default has set up to work in man-
aged/operational mode. Therefore, in manage mode setup client can connect to an access point through
SSID name and pre-shared password key which is the common way for all wireless devices to make
a connection from client to access point. Otherwise, in monitor mode wireless connection between
client and access point disconnect, this means that it is not possible for a client to have connection to
an access point when the wireless card is in monitor mode status. This is because monitor mode only
capture radio-layer information about packets. Therefore, connection between client and access point
terminates in that situation (discussion chapter explains more in detail about that problem). For the
mentioned reason, one Raspbery Pi’s wireless card is configured to work stand-alone in monitor mode
so therefore this card is responsible to capture data-link layer.

Steps presented in (3.1 and 3.2 below) required to set the wireless card to monitor mode in Kali
Linux which customized for Raspberry Pi. The reason why it requires to use this specific Kali Linux
distribution is that it can set the wireless card from manage mode to monitor mode through a specific

26 CHAPTER 3. APPROACH
 Approach

driver called NEXMON due to [30]. In addition, the command airmong-ng can be used to enable the
monitoring mode on that device in desired channel which requires to capture. Commands enabled
monitor mode in two different channels such as 108 and 140 in 5 GHz frequency band are as follow:
[31]

airmon-ng start wlan0 108 —— to observe traffic on channel 108 . (3.1)

airmon-ng start wlan0 140 —— to observe traffic on channel 140 . (3.2)

To switch from monitor to manage mode steps (3.3) and (3.4) are required.

airmon-ng stop wlan0mon 108 —— stop monitor mode capturing traffic on channel 108 . (3.3)

airmon-ng stop wlan0mon 140 —— stop monitor mode capturing traffic on channel 140 . (3.4)

When commands such as 3.1 and 3.2 are executed in the Raspebbry Pi who has a role of monitoring
the status of wlan0 interface changes to wlan0mon. Then, through this interface traffic can be captured
through Wireshark. Rasbperry Pi device who is reponsible to capture air traffic has always fix place in
the environment test were performed. The position of Monitor card has shown in figures 3.1 as well as
3.2.

The complete information about the difference between Manage mode and Monitor mode has been
discussed in the discussion chapter of the report.

3.2 Throughput testing

The purpose of doing experiments with a single and two pair of networks is to observe how many
frames transmits from client to access point, how many received by access point, the number of pkt
send by client and received by access point in different certain distances when the channel is busy and
when the channel is not in use. Therefore, when the Client 2 and AP2 work as an interfering network
beside Client1 and AP1 result is comparable for further analysis. This means to observe when channel is
corrupted by other nodes to measure the throughput and network performance. In addition, to observe
how RTS/CTS mechanism acts in a measurement when RTS is under control by access point and when
it is off. Moreover, how many frame loss produces when received signal in client is weak and strong
when a channel is busy or free. Also, to observe how CSMA/CA behaviour in a case channel is idle or
free to measure throughput when two nodes transmitting frame at the same time in a channel and how
they reduce the performance of network.

CHAPTER 3. APPROACH 27
Approach

3.2.1 The mechanisms of the throughput testing

The server application runs in server side which is access point. After execution it waits to answer to
the client request.

3.2.2 Theory of operation

The measure of successfulness of packet transmission is the throughput which is the number of bits
transmitted due to [32].

Prob success transmission ∗ Mega bits transmitted

Throughput = .
Time f or 1 try o f 1 packet to be sent on the air

The test was accomplished in a same time frame for the entire type of test which is 100 seconds. traffic
generated through iperf3 throughput tester application which is a well-known tool according to [9]. To
run the throughput testing commands is used on both sides on the server and on the client. So those
are as follows:

• Server: iperf3 -s -p 2323

• Client: iperf3 -u -c ”server IP address” -p 2323 -t 100

The server side command run iperf3 in server mode and it listens to port 2323, -s run iperf3 in server
mode. Command runs in client side specifies the UDP packet through -u command and -c specifies
the current machine is a client. In addition, -t decides for timeframe which test should be completed.
Therefore, tests setup to run in 100 seconds for the measurements. It is worth mentioning that when
those command runs iperf3 maximize the throughput by flooding the channel with 1470 Bytes long
UDP packet by default. In order to decide the length of the packet size the –length parameter is used to
observe the result when the size of the packet is higher than default. command are as follows:

• Client: iperf3 -u -c ”server IP address” -p 2323 –length 2100 -t 100

• Client: iperf3 -u -c ”server IP address” -p 2323 –length 6560 -t 100

Therefore, through the mentioned steps throughput testing has been conducted and result has shown
in the result chapter of the report.

channels such as 108 and 140 were assigned to the access points to work on those channels. Traffic sent
on the air on Client2 and AP2 network with the goal of flooding channel and making a pair of interfere
network for pair one network. In this situation (when two pairs of network working beside each other)
both APs working in the same channel, either 108 (busy) or 140 (free) in different tests. In addition, the
wireless monitor card work in monitor mode to capture data link layer information and record them
in a file. The number of pcap files are generated through Wireshark which sniffed data link layer traffic
during measurement.

One meter and five meter distances between client and access point has been considered for the first
and second measurement. Figure 3.1 and the subsequent table illustrates the measurement setup and

28 CHAPTER 3. APPROACH
 Approach

the configuration of devices used during measurement. According to the document of the hostapd [33]
supported rates are varied and it depends on which hardware can support what rate. The rate 240
equal to 24 Mbit/s showed it works stable in Raspberry Pi device after configuration was completed.
Therefore, rate fixed in 24 Mbit/s on both access points in configuration file of hostapd.

Figure 3.1: Single network, One and Five meter distances between client and access point.

RF interface type Broadcom Wi-Fi network interface (Raspberry Pi Wi-Fi interface)

Frequency 5540 and 5700 (5GHz)
Channel 108 and 140
Bitrare Fixed (24 MBit/s)
Nodes 2
Access Points 1
Client 1

Table 3.1: The measurement architecture with one network

CHAPTER 3. APPROACH 29
Approach

In the second measurement, the distance between client and access point increased by five meters. This
distance is the most far distance exist in the test environment (the authors apartment building). First
Laptop computer is connected through RJ-45 cable to the Raspberry Pi device who has role of monitor-
ing and Wireshark started to capture the air traffic. Then, the same cable unplugged from Laptop and
plugged to the Client1 device to start throughput testing on the channel. Server must start first before
client can start.

At this stage the measurement with single network in two different certain distances has completed and
the Monitor wireless card captured required data from air. The third measurement is the measurement
with the presence of the controlled interfering network which is Client2 and AP2. As figure 3.2 illustrates
not controlled neighboring access points are working in the same environment. To complete the test
first the Monitor wireless card started, then the traffic sent to the Client2 and AP2 network to flood
the channel in order to make an interfering network for Client1 and AP1. Finally, traffic sent on the
Client1 and AP1 network. Test has been accomplished in two different certain distances such as one and
five meter distances between two wireless networks (Client1 and AP1 and Client2 and AP2). Figure 3.2
shows two networks including two access points and two clients working beside each other in the same
test environment. Table illustrates the configuration on both devices in two different measurements in
channels 108 and 140.

Figure 3.2: Controlled and uncontrolled interfering networks are available and they are
working in the same environment. One and Five meter distances d (distance) between two
wireless networks.

30 CHAPTER 3. APPROACH
 Approach

RF interface type Broadcom Wi-Fi network interface (Raspberry Pi Wi-Fi interface)

Frequency 5540 and 5700 (5GHz)
Channel 108 and 140 in both APs
Bitrare Fixed (24 Mbit/s)
Nodes 4
Access Points 2
Client 2

Table 3.2: The measurement architecture with two networks

3.3 Tools used for capturing data

To do the measurement, the following tools has been used:

• Throughput tester application used during measurement to sent traffic on the air on both links.

• hostapd 1 has been used to configure access points with desired radio band and channel
frequency in the Raspberry Pies.

• Kali Linux operating system customized for Raspberry Pi device who is responsible to capture
air traffic. The Monitor mode device operating system 2

• Rasperian OS 32-bit Debian-based Raspberry Pi’s recommended OS on both clients and access
points.

• External RJ-45 connector located in Laptop to connect the raspberry Pi to the Laptop and access
to the console of Raspberry Pi devices.

• Ethernet cable to connect Laptop to the Raspberry Pi devices and read collected data.

• Wireshark to sniff the low level data frames in data link layer.

3.4 Wireshark Filtering feature

The number of frame transmitted to the access point has been measured through the Python script. In
addition, this script calculates how much data frame send to the desired access point in Byte. Transmit-
ted frames to access point observed and reported in the result chapter. It is possible to observe numbers
through Wireshark application as well. In order to access to Wireshark statistics a few steps should be
done. For instance, in Wireshark in the menu bar there is an item is called wireless, after clicking on
wireless item the WLAN Traffic items should be triggered. Therefore, it is possible to observe desired
statistics. For example, how many data packet sent from client to access point and how many received
through the based station (AP). In addition, it is possible to read the Retry bit number, the percentage
of re-transmitted frames and other statistics regarding specific measurements. The number of Retry bit
which is desired number for this project obtained form Wireshark statistics.

1 hostapd is a user space daemon software enabling a network interface card to act as an access point and
authentication server.
2 Kali-linux-2021.1-rpi4-nexmon-64 [34]

CHAPTER 3. APPROACH 31
Approach

3.5 Operating system of Raspberry Pie

Two Raspberry Pies configured as access point and they uses a Rasbian operating system which is
Debian-based Linux operating system. In addition, the both clients uses the same Linux operating
system as access points use during the measurement. The hostapd software used to make two Raspberry
Pis as access points. Through hostapd software the wireless network interface card is enabled to act as
access point and authentication server for the clients during the measurement. Kali Linux (Customized
for Raspberry Pi) is used to use the monitoring feature of WiFi interface card to observe traffic on the
air and to capture frames in data link layer.

3.5.1 Rasspbery Pi’s MAC address

Clients and access points Media Access Control addresses are shown below. Those address are required
to have access to related data captures by Monitor mode Wireless card. This is because to exclude traffic
traveled between

(dc : a6 : 32 : 32 : 40 : 45 ==> Client1) − (dc : a6 : 32 : 32 : 3 f : f a ==> AccessPoint1) (3.5)

(dc : a6 : 32 : 32 : 3 f : d0 ==> Client2) − (dc : a6 : 32 : 32 : 40 : 42 ==> AccessPoint2) (3.6)

32 CHAPTER 3. APPROACH
Chapter 4

Results and Analysis

This chapter explains the results of the measurements in which the testbed devices are used for the
throughput testing and performance evaluation. As explained in the approach chapter each measure-
ment has been carried out in two different distances; one meter is the closest distance possible and five
meter is the largest possible distance in the lab the experiments were performed.

It is important to mention that the Client2 and AP2 network has the role of interfering network for the
Client1 and AP1 network. Therefore, with the presence of the interfering network, distance between two
wireless networks is one and five meter(the horizontal distance).

The distance between clients and its access point is always one meter (the vertical distance) without the
presence of Client2 and AP2.

4.1 Range limitations

The tests were carried out in a small lab limited the range available for tests. The distances for the
interference measurements where limited to d1 = 1 and d2 = 5 meters distances between the two
interfering networks. The path loss in dB is given by

L = 10 · log10 (−d/λ)η (4.1)

where η is the propagation exponent and λ is the wavelength. The difference in path loss between these
two distances is therefore:
Ld = 10 · η log10 (−d1 /d2 ) (4.2)

For η = 2, corresponding to free space propagation, the difference of only 12 dB.

4.2 Result of experiments

Result of the experiment with two different channels such as 108 as well as 140 will be illustrated in a
form of figures and tables. In channel 108 there are a few access points working in the lab environment
which they are not controlled (meaning that we do not have access to them). Also, channel 140 was
observed free of use by other access points. The reason is that it was used two different channels is to

33
Result

observe different results. WiFi Analyzer mobile application was used to to observe channels.

Figures illustrates result of experiments without presence of controlled interfering network (only single
network with different pkt size). Tables demonstrates all results more in detail and exact numbers ac-
cumulated from measurement pcap file.

In order to observe the behavior of throughput with different pkt size, the length of the pkt increased
from default value ( 1470 bytes ) to 2100 and 6560 bytes in different measurements. The length decided
through TP tester application and explained in approach chapter through Linux command. In the mea-
surement when the pkt size 2100 (bytes) it was observed that the first frame was transmitted to its access
point has a length of 138 bytes. As it is not possible to transmit a packet more that 1500 bytes cause it
goes over maximum transmission unit size. Therefore, the TP tester fragmented pkt size to smaller size
and transmitted frames. Numbers such as 700 and 1500 were most likely observed as the length of
frames in the related pcap file.

To observe the length of frames a Wireshark filter is required. As Monitor card captures all wireless traffic
on the air (related or not related to the link you are working on). Therefore, a filter is required to exclude
frames transmitted to its access point. Therefore, filters were used to filter traffic are as follows: (4.4 is
used in addition to 4.3)

(wlan.addr == dc : a6 : 32 : 32 : 40 : 45)&&(wlan.addr == dc : a6 : 32 : 32 : 3 f : f a)&& (4.3)

[35]wlan. f c.types ubtype == 40) (4.4)

(wlan.addr == dc : a6 : 32 : 32 : 40 : 45) ==> Client − MAC − address (4.5)

(wlan.addr == dc : a6 : 32 : 32 : 3 f : f a) ==> AccessPoint − MAC − address (4.6)

As table 4.12 illustrates when the interfering network is not present beside Client 1 throughput is higher
than 100 MBit/s stay at 103.69 MBit/s. In contrast, in the same measurement when the interfering net-
work is present throughput is under 100 MBit/s stay at 89.29 MBit/s. ( Table 4.3 and 4.12 compares).
The reason is the channel they used is free of use without the presence of interfering network so there-
fore medium is idle since there is no access point working in the same channel. Therefore, throughput
without the presence of interfering network is better (table 4.12). In addition, distance between client
and access point without the presence of interfering network is one meter. So, received signal by client
is more reasonable in one meter distance. As Eric [5] claims the best signal is -30 dBm and least signal
is -90 dBm. In this measurement when we remove interfering network received signal by client is -46
dBm. Therefore, this is the second reason shows throughput is better.

As the introduction chapter explained the RTS/CTS is implemented in access point in conjunction with
CSMA/CA to reduce the risk of collision. Due to [8] RTS/CTS as a virtual career sensing is not a
complete solution to improve throughput and it may reduce throughput even further, but adaptive
acknowledgment from the base station can help to reduce the risk of re-transmission. This is because
client send request to send message to access point and ask if the medium is idle, then transmits frame,

34 CHAPTER 4. RESULTS AND ANALYSIS

 Result

but if the medium is busy then it wait until to receive CTS message from access point for transmis-
sion. Table 4.2 illustrates when the RTS mechanism is under control by access point the number of
re-transmission is under 5 stay at 4 on both clients. Also, the percentage of re-transmission is 0. In
contrast, in the measurement when the RTS mechanism is not controlled by access point the number of
re-transmission is higher than 200 stay at 223 number in client 1 and 538 in client 2. (Tables 4.2 and 4.3
compares).

Result in the measurement with larger pkt size shows a little bit higher throughput. For instance, in the
measurement when pkt size is 2100. This is because the larger pkt size do not need to have access to
the medium as the smaller pkt size required. This is because the smaller pkt size has more overhead
than larger so therefore it needs to have more access to the medium to send the same number of bytes
which the larger packet size send. Table 4.8 represents the transmitted frames is 14,082,704 bytes when
pkt size is 2100 compared with the measurement when the pkt size is 1470 the transmitted frames is
13,443,593 bytes ( tables 4.8 and 4.10 shows the matter). Therefore, throughput in the former is 112.66
MBit/s, while throughput in the latter is 107.54 MBit/s . During the TP testing when the pkt size was
decided to send with the length of 2100 there was a fragmentation occured through the TP tester ap-
plication. After investigating in the measurement file it was observed that frames transmitted to the
corresponding access point has a length of 1578 and 746 bytes with the same header length of 24 bytes.
As it it not possible to transmit a packet more than 1500 byte since MTU size is 1518 bytes (1500 packet
and 18 bytes for header) due to [36] . Then, fragmentation occured in application layer. Therefore, the
fragmented length of packet was observed in data link layer in pcaps. .

Tables 4.4 illustrates throughput when two wireless networks have one meter distance from each other
is under 100 MBit/s stay at 89.92 MBit/s in Client 1 and 86.82 MBit/s in client 2. In contrast, when they
have five meter distance from each other they have a better throughput which is 106.32 MBit/s in Client
1 and 105.31 MBits/ in client 2 (table 4.5). This is noticeable that as they have a one meter distance from
each other they should shared medium in halve equal so therefore 50 percent reduction in throughput
is expected, but result shows only 10 to 20 percent reduction in throughput. A few probability are exists
in this condition. One reason is that there is a little bit delay starting the TP testing in Client 1. This
is because it was one Laptop and one cable available and it required to switch between three different
devices (Two networks (Client1 and Client2) as well as one monitor card) at the same time to start TP
testing. Therefore, it could mention that channel is not busy for at least 5 - 10 seconds in the beginning
of test.

Result shows when distance increases gradually throughput goes down increasingly. The reason is
the power transmitted by access point decreases in larger distance. Therefore, RSSI or received signal
from client reduces subsequently and result in the lower throughput and re-transmission of frames.
Therefore, throughput form 103.69 MBit/s in one meter distance reaches to 88.46 MBit/s in five meter
distance. This would be a one example explained in problem statement which claims when signal is not
detected transmission start and result in re-transmission and more Retry bit . Moreover, the percentage of
re-transmission goes up from 1.2 to 2.9 percent ( tables 4.12 and 4.13 demonstrates the matter). In ad-
dition, in the measurement when pkt size is 2100 when client and access point have one meter distance
throughput is 112.66 MBit/s a little bit above 103.69 MBit/s (in pkt size 1470 shows in table 4.12), while
distance increases and reaches to five meter throughput goes down reaches at 95.10 MBit/s this is also
a bit over 88.46 MBit/s (in pkt size 1470 table 4.13). Tables 4.8 and 4.9 illustrate when pkt size is 2100.

CHAPTER 4. RESULTS AND ANALYSIS 35

Result

As the introduction chapter explained regarding career sensing with collision avoidance for wireless
network figure in 1.3, it was observed and analysed when channel is busy by interfering network, and
RTS is not under control by access point throughput is low. This can be shown in table 4.4. Also, inter-
ference is high in one meter distance between two wireless networks as well. On the other hand, results
shows the same measurement when distance is five meter between networks throughput is better (table
4.5). This is because two networks are far away from each other and interference is less since channel
is not more in use. In addition, result shows when RTS in under control by access point throughput is
109,26 and 103.89 MBit/s in Client 1 and Client 2 (Table 4.1). Compared with the measurement when
the RTS is not under control throughput is 89,29 and 99,41 MBit/s in Client 1 and Client 2 respec-
tively in one meter distance. Tables such as 4.1 and 4.3 illustrate the matter. Also, the percentage of
re-transmission is 2.1 and 0.1 in Client-1 and Client-2 when the RTS is under control compared with
the measurement when RTS is not under control re-transmission percentage is 2.8 and 6.4 in Client-1
and Client-2 (Tables such as 4.1 and 4.3) shows the matter. Also, it is possible to refer to the Jamieson
[20] et al. which they claim that ”when interference is local and nodes are within each other’s transmis-
sion range, carrier sense and RTS/CTS exchange may be a good method of contending for the channel”.

In single network measurements without the presence of interfering network with different pkt size re-
sults represented when the pkt size is larger re-transmission is higher compared with the measurement
when the pkt size is smaller. Tables such as 4.8 and 4.12 illustrates the matter. This is because the smaller
pkt needs more access to medium than larger packet therefore less transmission occurs when pkt size
is smaller. On the other hand, larger pkt has fewer overhead so less access to medium is required there-
fore more frames transmits in bytes. But it is noticeable that different measurement with different pkt
size represents different results. For instance, in a measurement when pkt size decided 2100 without
presence of interfering network the percentage of re-transmission is over 4.5 stay at 4.7 percent (shows
in table 4.8). But in the measurement with the same distance when the pkt size is 1470 the percentage
of re-transmission is under 0.5 stay at 0.4. Tables such as 4.8 and 4.16 represents the matter. This is
because it takes more time for larger pkt to be sent than the smaller when they are transmitting in the
same time-frame.

The RTS/CTS feature in the measurements without presence of interfering network shows different
results. For instance, Tables such as 4.11 and 4.16 shows although RTS is under control by access point
in the former (table 4.11), but the percentage of re-transmission is higher compared with the measure-
ment when the RTS feature is not under control by access point. Therefore, the former has 7.4 (table
4.11) percent of re-transmission and the latter has the 0.4 percentage of re-transmission. This is because
RTS generally tries to solve the hidden node problem when there are more than one network send a
message to access point. This is because to control clients when they try send messages to access point
and expect to receive CTS with the goal of reducing collision.

In order to observe frame size in the measurement with pkt size 6560 the same steps (introduced above)


are required in the related pcap file. After execution of the mentioned filter 4.3 and 4.4 it was observed
that the frames transmitted to the corresponding access point has a length of 1578 and 746 bytes with
the same header length of 24 bytes. As explained above fragmentation occured and this is because MTU
size is 1500 bytes. Therefore, frames should be transmitted in smaller amount of numbers.

36 CHAPTER 4. RESULTS AND ANALYSIS

 Result

Figure 4.1 illustrates throughput when Client1 and AP1 work stand-alone without the presence of Client2
and AP2. The range limitation calculated when Client1 and AP1 work stand-alone. Signal in five meter
distance is -58 (dBm), this number deducted with signal accumulated in one meter distance (-46 (dBm)).
Therefore, the difference is 12 db. This calculation conducted based on formula (4.1 and 4.2) presented
above.

As figure 4.1 illustrates when distance between client and access point is one meter throughput is better
compared with the other side when client and access point has five meter distance from each other.
This is because in one meter distance received signal from access point is stronger and in five meter
distance received signal is weaker. But as distance is not exceeds more than five meter there is not a big
differences in one and five meter distances in throughput variation. In this measurement the controlled
interfering network is not present during measurement. The received signal problem introduced in
problem statement was showed and compared through different measurements in table format above.

Figure 4.2 illustrates the percentage of re-transmitted frames to the corresponding access point with
different pkt size. In this measurement the controlled interfering network is not present during
measurement. This probability calculated through below formula:

100 / [pkt sent] * retry bit (4.7)

In order to calculate this probability the number of Retry bit is required. It is worth nothing that the Retry
bit number observed through Monitor Mode wireless card after measurement was completed through
pcap file. Thus, for this purpose all frames re-transmitted should be divided by 100 (this is because all
frames are 100 percent of the frames). Then, the accumulated number should be multiplied by Retry bit
number. As figure 4.2 illustrates there are variations on re-transmission with different pkt size. The per-
centage of re-transmission with larger pkt size should be higher than the percentage of re-transmission
with smaller pkt size since the larger pkt need less medium access than smaller then the number of
Retry bit goes up accordingly. But there is a question why pkt size 6560 has less Retry in this specific
measurement. Figure 4.3 shows the number of Retry instead of percentage .

Table 4.4 and 4.5 illustrates result of throughput testing when they have one and five meter distance
from each other. Based on those statistics throughput when two wireless networks have one meter dis-
tance from each other is lower compared with the measurement when they have five meter distance.
This is because in one meter distance interference is high so therefore rate goes down and result in a
lower throughput and re-transmission number increases.

Table 4.6 illustrates result of measurement when access points have setup to work in channel 108 which
was observed busy in the test environment, but not so busy. It was observed that there are five other
not controlled access points working in the same channel (108). In this measurement the pkt length has
decided through TP tester application to be 6560 to observe result when packet has larger size than de-
fault size (1470). In this measurement two wireless networks have one meter distance from each other.
It is interesting to mention that in this measurement when two wireless networks are far away from ech
other they have a better performance shows in table 4.7. For instance, throughput in one meter distance
has lower value stay at 87.42 MBit/s in Client 1 as interfering by Client2 and AP2, while when they have
five meter distance throughput is over 100 stay at 105.42 MBit/s table 4.7 represents. This is because

CHAPTER 4. RESULTS AND ANALYSIS 37

Result

the fewer interfering are exist in five meter distance between them.

Table 4.10 and 4.11 illustrates result when only Client1 and AP1 network works stand-alone in one and
five meter distance. In this measurement access point setup up to work in channel 108 which was ob-
served that four other not controlled access points working on that channel near of test environment.
As results represents in this measurements the percentage of re-transmission in the five meter distance
is over 7 percent stay at 7.4 in table 4.11. In contrast, in one meter distance as table 4.10 illustrates the
percentage of re-transmission is under 1.5 percents stay at 1.3. One reason is a distance. As distance
increases received signal goes increasingly weaker and the number of re-transmission goes up. This is
one example of measurement which can be refer to the problem statement introduced in introduction
chapter. ”if no signal is detected, transmission starts, This would result in many faults in the frames and also
result in re-transmission of frames.”

Tables 4.12 and 4.13 are result of measurements when only one wireless network work stand-alone,
but in channel 140 which was observed free of use by not controlled nearby access points in the test
environment. As channel is free of use it is possible to mention that the re-transmission percentage is
lower compared with the measurement when the channel is in use by other access points this is because
medium is more in idle when channel is free of use.

Tables 4.16 and 4.17 are repeat measurement. The reason of doing same measurement is to observe the
different result and the behaviour of the network.

Figure 4.1: Throughput without the presence of interfere network in two different distances.

38 CHAPTER 4. RESULTS AND ANALYSIS

 Result

Figure 4.2: The percentage of frames re-transmitted without the presence of interfere network
in two different distances.

Figure 4.3: Frame re-transmitted without the presence of interfere network in two different
distances.

CHAPTER 4. RESULTS AND ANALYSIS 39

Result

Items Client 1 Client 2

Throughput (MBit/s) 109.26 103.89
Transmitted frames (bytes) 13,658,141 12,987,469
Pkts Sent(Cli) 9367 8966
Pkts Received (AP) 8531 8157
Retry 183 6
Percent retry 2.1 0.1

Table 4.1: Result of experiment when two wireless networks have One meter distance from
each other. Channel 108 pkt size 1470. Frequency 5540MHz (Band 5GHz). RTS threshold 2346
bytes.

Items Client 1 Client 2

Throughput (MBit/s) 103.59 106.06
Transmitted frames (bytes) 12,949,458 13,258,045
Pkts Sent(Cli) 8936 9056
Pkts Received(AP) 8792 8931
Retry 4 4
Percent retry 0.0 0.0

Table 4.2: Result of experiment when two wireless networks have Five meter distance from
each other. Channel 108 pkt size 1470. Frequency 5540MHz (Band 5GHz). RTS threshold 2346
bytes.

Items Client 1 Client 2

Throughput (MBit/s) 89.29 99.41
Transmitted frames (bytes) 11,161,991 12,426,424
Pkts Sent(Cli) 7701 8420
Pkts Received(AP) 7635 8336
Retry 223 538
Percent retry 2.8 6.4

Table 4.3: Result of experiment when two wireless networks have One meter distance from
each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz). RTS is Off.

40 CHAPTER 4. RESULTS AND ANALYSIS

 Result

Items Client 1 Client 2

Throughput (MBit/s) 89.92 86.82
Transmitted frames (bytes) 11,239,696 10,851,960
Pkts Sent(Cli) 7901 7520
Pkts Received(AP) 7741 7361
Retry 196 154
Percent retry 2.4 2.1

Table 4.4: Result of experiment when two wireless networks have One meter distance from
each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz). RTS is Off. Repeated
measurement of table 4.3.

Items Client 1 Client 2

Throughput (MBit/s) 106.32 105.31
Transmitted frames (bytes) 13,290,610 13,164,902
Pkts Sent(Cli) 9239 8973
Pkts Received(AP) 9099 8868
Retry 7 238
Percent retry 0.1 2.7

Table 4.5: Result of experiment when two wireless networks have Five meter distance from
each other. Channel 140 pkt size 1470. Frequency 5700MHz (Band 5GHz). RTS is Off.

Items Client 1 Client 2

Throughput (MBit/s) 87.42 119.44
Transmitted frames (bytes) 10,928,408 14,930,332
Pkts Sent(Cli) 8183 10036
Pkts Received(AP) 8076 9945
Retry 116 31
Percent retry 1.4 0.3

Table 4.6: Result of experiment when two wireless networks have One meter distance from
each other. Channel 108 - pkt size 6560. Frequency 5540MHz (Band 5GHz). RTS is Off.

CHAPTER 4. RESULTS AND ANALYSIS 41

Result

Items Client 1 Client 2

Throughput (MBit/s) 105.42 149.35
Transmitted frames (bytes) 13,178,081 18,669,650
Pkts Sent(CLi) 9821 13833
Pkts Received(AP) 9737 13744
Retry 263 190
Percent retry 2.7 1.4

Table 4.7: Result of experiment when two wireless networks have Five meter distance from
each other. Channel 108. Pkt size 6560. Frequency 5540MHz (Band 5GHz). RTS is Off.

Items Client 1
Throughput (MBit/s) 112.66
Transmitted frames (bytes) 14,082,704
Pkts Sent(CLi) 12860
Pkts Received(AP) 12790
Retry 608
Percent retry 4.7
RTS threshold (bytes) 2346

Table 4.8: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 2100. Frequency
5540MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 95.10
Transmitted frames (bytes) 11,887,647
Pkts Sent(CLi) 10826
Pkts Received(AP) 10785
Retry 806
Percent retry 7.4
RTS threshold (bytes) 2346

Table 4.9: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 2100. Frequency
5540MHz (Band 5GHz).

42 CHAPTER 4. RESULTS AND ANALYSIS

 Result

Items Client 1
Throughput (MBit/s) 107.54
Transmitted frames (bytes) 13,443,593
Pkts Sent(CLi) 9144
Pkts Received(AP) 9046
Retry 120
Percent retry 1.3
RTS threshold (bytes) 2346

Table 4.10: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 1470. Frequency
5540MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 100.24
Transmitted frames (bytes) 12,530,199
Pkts Sent(CLi) 8521
Pkts Received(AP) 8406
Retry 626
Percent retry 7.4
RTS threshold (bytes) 2346

Table 4.11: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 1470. Frequency
5540MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 103.69
Transmitted frames (bytes) 12,961,709
Pkts Sent(CLi) 8763
Pkts Received(AP) 8653
Retry 102
Percent retry 1.2
RTS/CTS Off

Table 4.12: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 140. Pkt size 1470. Frequency
5700MHz (Band 5GHz).

CHAPTER 4. RESULTS AND ANALYSIS 43

Result

Items Client 1
Throughput (MBit/s) 88.46
Transmitted frames (bytes) 11,058,450
Pkts Sent(CLi) 7470
Pkts Received(AP) 7383
Retry 216
Percent retry 2.9
RTS/CTS Off

Table 4.13: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 140. Pkt size 1470. Frequency
5700MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 105.41
Transmitted frames (bytes) 13,177,100
Pkts Sent(CLi) 9822
Pkts Received(AP) 9719
Retry 165
Percent retry 1.7
RTS/CTS Off

Table 4.14: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 6560. Frequency
5540MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 45.85
Transmitted frames (bytes) 5,731,370
Pkts Sent(CLi) 4287
Pkts Received(AP) 4232
Retry 151
Percent retry 3.6
RTS/CTS Off

Table 4.15: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 6560. Frequency
5540MHz (Band 5GHz).

44 CHAPTER 4. RESULTS AND ANALYSIS

 Result

Items Client 1
Throughput (MBit/s) 108.62
Transmitted frames (bytes) 13,577,109
Pkts Sent(CLi) 9237
Pkts Received(AP) 9085
Retry 32
Percent retry 0.4
RTS/CTS Off

Table 4.16: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is One meter. Channel 108. Pkt size 1470. Frequency
5540MHz (Band 5GHz).

Items Client 1
Throughput (MBit/s) 28.98
Transmitted frames (bytes) 3,621,915
Pkts Sent(CLi) 2667
Pkts Received(AP) 2441
Retry 21
Percent retry 0.8
RTS/CTS Off

Table 4.17: Result of experiment when only Client1 and AP1 wireless network works stand-
alone. Distance between client and AP is Five meter. Channel 108. Pkt size 1470. Frequency
5540MHz (Band 5GHz).

CHAPTER 4. RESULTS AND ANALYSIS 45

Result

and Analysis

46 CHAPTER 4. RESULTS AND ANALYSIS

Chapter 5

Discussion and Future Work

In this chapter measurement tools, actions required to do the measurements as well as difficulties and
challenges occured during measurements will be discussed. In addition, any probable weakness of
the measurement tools and methodology was used for the measurements will be discussed.Issues
regarding hardware and software, possible adjustment as well as possible future works will be
discussed at the end.

5.1 Problems and challenges occured during measurement

As approach chapter explained the Raspberry Pi devices with a common Debian-based Rasberian op-
erating (32-bit) system was used for client machines during the measurements. By default wireless net-
work interface card is setup to work in promiscuous mode or manage mode. Manage mode is used to
make a wireless link to its corresponding access point. Therefore, it is not possible to sniff the data link
layer’s behaviour when wireless card is in promiscuous mode. It required to understand how it is pos-
sible. By this research study the author realized that in order for capture MAC frame fields and access
to the required items it is required to set the wireless network interface in the Monitor mode. Therefore, it
enables the wlan interface to capture/sniff data link layer frames and subsequently to understand how
many frames transmitted at once how many re-transmitted. Also, it gives in detail information about
WLAN behavior.

Due to [37] in order to capture radio-layer information about packets and observe information in this
layer wireless card must change from the state of manage mode to monitor mode. To address the prob-
lem the author investigated to understand how this work is possible. Since it is not possible for the
Raspberry Pi operating system (Which is recommended by default) to set the card from promiscuous
mode to monitor mode other operating system who can support the monitor mode feature was re-
quired for the project instead. Thus, after so many research and challenging the author was addressed
the problem through an specific kali Linux operating system customized for Raspberry Pies uses NEX-
MON firmware. As this firmware is integrated to the kernel of the kali Linux operating system cus-
tomized for Raspberry Pi it doesn’t require to install it afterwards. Due to [30] NEXMON is a C-based
firmware patching framework for Broadcom/Cypress Wi-Fi chips enables Monitor mode with radiotap
headers and frame injection.

After setting the card from promiscuous mode to monitor mode it was observed that the card in some

47
Discussion

measurement do not work 100 percent stable. In this condition, as the card presented it is active, but
there was not traffic captured by the card so therefore a reboot for the entire operating system was re-
quired to repeat the measurement again.

The second challenge: When the Wi-Fi card is in Monitor mode it is not possible to keep client’s con-
nection to the corresponding access point. Therefore, it was required to realize that how it is possible to
keep client’s connection to access point and have access to Monitor feature to capture air traffic at the
same time. It is not possible to have access to both features promiscuous mode to monitor mode at the
same time. Therefore, the author had to come up with a solution. To address the problem following
alternatives has been considered:

Alternative one: To have an extra USB-based wireless card and make a bridge connection with Rasp-
berry pi’s wireless card. Therefore, one card can be in Monitor mode and can capture MAC frame fields
and the second card in Manage mode can keep the connection to the access point. Therefore, traffic can
be sent on the air and can capture frames from data link layer. This alternative was not successful. The
reason was the USB-based card could not be detected though Raspberry Pies OS after so much chal-
lenge. Therefore, other solution required to be considered. Figure 5.1 represents the first architecture
which was not successful.

Figure 5.1: Single network with USB wireless card on a client

48 CHAPTER 5. DISCUSSION AND FUTURE WORK

 Discussion

The second alternative: The second alternative has a different design architecture compared with
alternative one. In this solution one client and one access point can make a link through
manage/operational mode card. In addition, one Raspberry Pi device can work stand-alone in Monitor
mode and can capture radio-layer information. Also, it captures nearby (uncontrolled) wireless
network traffic on the air. Figure 5.2 illustrates the second alternative design. This solution was
successfully implemented and the desired fields captured through Wireshark and subsequently the pcap
files produced as a result of measurements.

Figure 5.2: Single network and monitor mode device captures traffic information from air.

CHAPTER 5. DISCUSSION AND FUTURE WORK 49

Discussion

5.1.1 Equipment problems

It is important to note that with the presence of the second network Client2 and AP2 as interfere network
some difficulties were experienced. For instance, it was required to do actions in three different devices
at the same time to start the test. So, those are as follows:

• Step one is to plug the Ethernet cable to Raspberry Pi’s Ethernet interface who has a role of
Monitor. Then, start Wireshark and push to start button to start capturing/sniffing air traffic. (This
is the place where that it was observed card do not work stable sometimes, therefore reboot in OS
and re-configuration on wlan card was required.)

• Step two is to unplug Ethernet cable from Monitor Raspberry Pi device and plug the cable to the
Client 2 and start throughput testing in order to send traffic on the link to flood the channel with
the purpose of making an interfere network for Client1 and AP1.

• Step three is to unplug the cable from Client 2 device and plug to Client 1 to start throughput
testing on Client1 and AP1.

All actions introduced above were required to execute at the same time as fast as possible in order to
run test in the same timeframe. Also, only one Ethernet cable was available to use and switch between
devices. Therefore, switching between devices to have access to the console of devices and do the action
was another challenge which made a bit delay to have access to the console of each device (a delay of 5
to 10 seconds). This is because it took a bit time to connect to the console of device after cable unplugged
from one and plugged to another device. It is better to have three Laptops computer so each Laptop
can control one device. Figure 5.3 illustrates the steps.

50 CHAPTER 5. DISCUSSION AND FUTURE WORK

 Discussion

Figure 5.3: Steps to capture air traffic with the presence of Client2 ans AP2 as a interfere network.

It is noticeable that the Raspberry Pi device who has a role of capturing data link layer, the Monitor


mode interface showed that it is not working stable and it crashes sometimes during measurements.
This is because it seems capturing/ sniffing air traffic is a very heavy process. Therefore, it requires
a professional wireless network interface to handle the work (Raspberry Pi Wi-Fi card do not recom-
mended). This is because after each measurement the operating system of Raspberry Pi required reboot.

Figure 5.4 represents steps was required to run TP testing by one Ethernet cable.

5.1.2 The measurement tool used before iperf3

Before using iperf3 for throughput testing, the other TP testing application was used. After doing
so many tests, the result showed that the TP tester is not suitable for measurement as it sends data
frame from client to access point and reverse back to client and it is not possible to decide for time of
throughput testing as well as Pkt size. In addition, it is not possible to control time when throughput
testing start. Therefore, this work should be done manually. In order to have control to measurements
it was required to go to more professional TP tester. To figure out the matter, iperf3 was used. iperf3 can
cover the required features and it was used by other researchers for throughput testing. For example
Michael et al [9].

CHAPTER 5. DISCUSSION AND FUTURE WORK 51

Discussion

Figure 5.4: Steps to capture air traffic without the presence of Client2 and AP2 as a interfere
network.

5.2 Result of measurement accumulated by other researcher

For experiment, Michael et al [9] used their own platform and devices which is different with the de-
vices was used in this project. According to Michael they used Laptops (meaning that more than one)
with Netgear WAG511v2 PCMCIA cards 802.11 a/b/g (AR5212 chipset and the Madwifi driver). In
addition, they claim that they used CORAL testbed which is pretty specific so the refer explain that
testbed. CORAL [38] and [39] is developed by the Communication Research Center (CRC) to provide a
reseach study platform for studying interference in WLAN 802.11 b/g networks according to [9]. They
claim that the platform itself contains two different parts: a hardware called WIFI-CR and a software for
storing obtained interference information and controlling the hardware. The mentioned that the WIFI-
CR contains of a common off-the-shelf roruterboard RB433 from Miktotik which is equipped with two
Wistron CM9 802.11 a/b/g miniPC cards (AR5212 chipset). The claim that the first card is set to promis-
cuous mode and is used to sniff 802.11 b/g packets independently of the second transceiver card. The
routerboard is encapsulated by a shell of FPGA and RAdio Front-end (RF) circuitry to control the traffic
flows of the routerboard. Also, They mention the shell involves a simple built-in spectrum analyzer
for energy detection. So, in this project the author was use WiFi Analyzer which was downloaded form
Google play and was the best possible tool available to detect nearby access point and see and analyse

52 CHAPTER 5. DISCUSSION AND FUTURE WORK

 Discussion

in which channel they work. This work was required to find a most clear channel available in the test
environment. They claim the operating system of the routerboard is OpenWRT. Also, they mention
that the drawback of the encapsulation approach is a decrease of the maximum throughput by at least
20 percent. Also, only the 2.4 GHz band can be used.

It is noticeable that they have access to the distances form 0 to 18 meter for doing measurements in
their Lab and they compare the different throughput result in different distances such as 4, 8, 12, 15 and
18 meters. To summarize the discussion regarding their research work, the result they accumulated is
different from result accomplished for this research project as they use different setup, wireless assets
as well as the lab environment they used. To detail information refer to [9].

5.3 Future work

For the future work of this project it is recommend to use more professional wireless assets rather than
Raspberry Pi devices. For instance, a wireless card can support monitor feature without the help of
operating system and a specific firmware in order to be able to handle the heavy process itself and
record data in its local memory card. Therefore, result would be more concrete and reliable. As it was
introduced above assets which Michael et al [9] used is more professional than Raspberry Pi devices
such as finding a reasonable OS to capture data link layer. In addition, distance is another factor which
should be take it into account. The longer distance the more diverse results can be obtained for further
analysis.

5.4 Challenges occured during special days

It is important to mention that working on the project entirely carried out in the author’s small flat
which limited the distance variation and affected other things. One reason is that it faced with the
pandemic restriction so that the author had to stay home and work alone entirely with the given
wireless assets in hand. Therefore, it made the work challenging and increased difficulties for the entire
work. Also, specifically when the author had to fix technical problems and challenge happened in
devices. Although he supported by his supervisor digitally, but the author believe working in the real
lab with more professional devices could be more interesting and reduces challenges specifically in the
pandemic days. By the way, the author worked so hard during the difficult days and he tried to produce
a project as perfect as possible with the help of his supervisors. The author believe sharing difficulties
happened during the work has positive effect on the future work of the project.

CHAPTER 5. DISCUSSION AND FUTURE WORK 53

Conclusion

54 CHAPTER 5. DISCUSSION AND FUTURE WORK

Chapter 6

Conclusion

The problem statement for this thesis explained, in IEEE 802.11 accessing to a wireless channel is con-
trolled by the carrier sense mechanism. Also, explained CCA is a function works in conjunction with
CSMA/CA to perform the channel access by observing the channel with the purpose of frame trans-
mission.

In this thesis we studied and observed when channel is busy it waits so therefore throughput goes
down and result in a low performance of wireless network. The mentioned result was observed and il-
lustrated through different measurements. For instance, when distance between two wireless networks
is one meter interference occurs and throughput goes down . In addition, the number of re-transmission
increases and produce faults and error. Moreover, we observed when the RTS mechanism in under con-
trol by access point a few percentage of re-transmissions occurs.

The second problem was explained in problem statement was when the received signal by client is
weak or not detected by client and re-transmission start. Then, the percentage of re-transmission in-
creases so therefore throughput decreases accordingly and result in a weak performance on that net-
work. Through different experiments illustrated that this problem occurs specifically when there is only
single wireless network works stand-alone and when distance between a client and it corresponding
access point increase.

It is noticeable that a few experiments are exist which they repeated with the purpose of observing
different results for further analysis and future work of this project. Also, the interesting aspect of this
work is that it reports network behaviour in the data link layer which is accumulated by the author.
It makes possible to observe directly number of re-transmission and percentage of re-transmission of
frames in different measurements.

Overall, it is noticeable that the wireless equipment’s, measurement tools as well as the environment
tests has been carried out are different from equipment’s that other researchers used in their research
project. This problem was discussed in the discussion chapter in detail. Therefore, in the measurements
with the presence of interfering network results showed surprisingly small degradation in throughput
compared with other measurements reported in [9].

55
Appendix

56 CHAPTER 6. CONCLUSION
Chapter 7

Appendix

7.1 The Python script

When the measurement of throughput testing has been carried out the pcap will be generated through
Wireshark in Monitor mode card. Therefore, pcap files are calculated to realize that how many data
frames transmitted from client to access points and the number of frames. It is possible to obtain the
same statistic from Wireshark as well. Written by Madeleine.

import argparse
import os
from time import sleep
import sys
from scapy.utils import PcapReader
from scapy.layers.dot11 import *
from scapy.packet import Packet
from scapy.all import *

def get_data_frames():
myreader = PcapReader(FILE_NAME)
for pkt in myreader:
if Dot11 in pkt and pkt[Dot11].type == 2:
yield pkt

def get_data_frames_from_to(fromNode, toNode):

for pkt in get_data_frames():
DS = pkt[Dot11].FCfield & 0x3
to_DS = DS & 0x1 != 0
from_DS = DS & 0x2 != 0

# If data frame going is going from Client ---> AP:

if to_DS and not from_DS and pkt[Dot11].addr3 == toNode and pkt[Dot11].addr2 == fromNode:
yield pkt

57
Appendix

def count_all_data_frames():
count = 0
for pkt in get_data_frames():
count = count + 1
return count

def count_data_frames_from_to(fromNode, toNode):

count = 0
frames = get_data_frames_from_to(fromNode, toNode)
for frame in frames:
count = count + 1
return count

def count_all_retry_data_frames():
count = count_all_data_frames()
retryCount = 0

if count == 0:
return 0

for pkt in get_data_frames():

retry_bit = pkt[Dot11].FCfield & 0x8!= 0
if retry_bit:
retryCount = retryCount + 1

if retryCount == 0:
return 0
else:
return retryCount / count

def count_retry_data_frames_from_to(fromNode, toNode):

count = count_data_frames_from_to(fromNode, toNode)
frames = get_data_frames_from_to(fromNode, toNode)
retryCount = 0

if count == 0:
return 0

for pkt in frames:

retry_bit = pkt[Dot11].FCfield & 0x8!= 0
if retry_bit:
retryCount = retryCount + 1

if retryCount == 0:
return 0

58 CHAPTER 7. APPENDIX
 Appendix

else:
return retryCount / count

def data_frame_bytes_from_to(fromNode, toNode):

count_bytes = 0
frames = get_data_frames_from_to(fromNode, toNode)
for pkt in frames:
count_bytes = count_bytes + len(pkt[Dot11].payload)
return count_bytes

def process_pcap(AP, CLIENT):

print("All data frames:", count_all_data_frames())
print("Data frames", CLIENT, '(client) ---->', AP, '(AP):',
print("\nPercentage of re-transmitted data frames:"
print("Percentage of re-transmitted data frames", CLIENT, '(client) ---->', AP,
print("\nData frame bytes", CLIENT, '(client) ---->', AP,

if __name__ == '__main__':
parser = argparse.ArgumentParser(description='PCAP reader')
parser.add_argument('--pcap', metavar='<pcap file name>',
help='pcap file to parse', required=True)
parser.add_argument('--client', metavar='<client MAC address>',
help='The transmitting node', required=True)
parser.add_argument('--ap', metavar='<AP MAC address>',
help='The receiving node', required=True)
args = parser.parse_args()

global FILE_NAME

FILE_NAME = args.pcap
AP = args.ap
CLIENT = args.client

if not os.path.isfile(FILE_NAME):
print('"{}" does not exist'.format(FILE_NAME))
sys.exit(-1)

print("Starting processing of PCAP file", FILE_NAME, "\n")

process_pcap(AP, CLIENT)
sys.exit(0)

CHAPTER 7. APPENDIX 59
Appendix

7.2 Wireshark filters

wlan. f c.types ubtype == 40)− > to f ilter QoS Data f rames. (7.1)

wlan. f c.types ubtype == 2)− > to f ilter all Data f rames. (7.2)

wlan. f c.tods == 1)− > to f ilter retry bit towards access point. (7.3)

wlan. f c. f romds == 1)− > to f ilter retry bit f rom access point to client. (7.4)

[35]

60 CHAPTER 7. APPENDIX
Bibliography

[1] Wi-Fi-Alliance. Wi-fi alliance® celebrates 20-year milestone, industry momentum continues.
[Online]. Available: https://www.wi-fi.org/news-events/newsroom/wi-fi-in-2019

[2] Wi-FiAlliance. Wi-fi alliance® wi-fi® predictions for 2021. [Online]. Available: https:
//www.wi-fi.org/news-events/newsroom/wi-fi-alliance-wi-fi-predictions-for-2021

[3] maui communications. Wifi channel description. [Online]. Available: https://www.

maui-communications.net/802-11-channel-frequency-allocation

[4] T. O’Brien. Channel planning best practices for better wi-fi. [Online].
Available: https://www.ekahau.com/blog/channel-planning-best-practices-for-better-wi-fi/#:
∼:text=The%202.4GHz%20band%20is,20%20MHz%2D160%20MHz).

[5] E. Geier. signal to noise. [Online]. Available: https://www.networkworld.com/article/2215287/

coping-with-wi-fi-s-biggest-problem-interference.html

[6] WIKI. Snr. [Online]. Available: https://en.wikipedia.org/wiki/Signal-to-noise ratio

[7] wiki. Csma/cd. [Online]. Available: https://en.wikipedia.org/wiki/Carrier-sense multiple

access with collision detection

[8] Wiki. Hidden node. [Online]. Available: https://en.wikipedia.org/wiki/Hidden node problem

[9] M. Doering, Ł. Budzisz, D. Willkomm, and A. Wolisz, “About the practicality of using partially
overlapping channels in ieee 802.11 b/g networks,” in 2013 IEEE International Conference on
Communications (ICC). IEEE, 2013, pp. 5110–5114.

[10] wiki PIFS. Factor affecting wifi performance. [Online]. Available: https://www.fastmetrics.com/
blog/wifi/factors-affecting-wifi-performance/

[11] S. Feirer and T. Sauter, “Seamless handover in industrial wlan using ieee 802.11k,” in 2017 IEEE
26th International Symposium on Industrial Electronics (ISIE), 2017, pp. 1234–1239.

[12] S. Woo and H. Kim, “An empirical interference modeling for link reliability assessment in wireless
networks,” IEEE/ACM Transactions on Networking, vol. 21, no. 1, pp. 272–285, 2012.

[13] T. Huehn, R. Merz, and C. Sengul, “Joint transmission rate, power, and carrier sense settings: An
initial measurement study,” in 2010 Fifth IEEE Workshop on Wireless Mesh Networks. IEEE, 2010,
pp. 1–6.

[14] P. Riihikallio. power level. [Online]. Available: https://metis.fi/en/2017/10/txpower/#:∼:

text=By%20default%20almost%20all%20WiFi,a%20fraction%20of%20the%20maximum.

61
Appendix

[15] NetSpot. Sir. [Online]. Available: https://www.netspotapp.com/help/troubleshooting-sir/#:

∼:text=The%20signal%2Dto%2Dinterference%20ratio,present%20from%20other%20radio%

20transmitters.

[16] “The effective of adjacent channel rejection and adjacent channel interference,” in International
Conference on Wired/Wireless Internet Communication. Texas-Instruments, 2003, p. 3.

[17] R. Gummadi, D. Wetherall, B. Greenstein, and S. Seshan, “Understanding and mitigating the
impact of rf interference on 802.11 networks,” ACM SIGCOMM Computer Communication Review,
vol. 37, no. 4, pp. 385–396, 2007.

[18] F. Metrics. affect interference. [Online]. Available: https://www.fastmetrics.com/blog/wifi/

factors-affecting-wifi-performance/

[19] A. Saif, M. Othman, S. Subramaniam, and N. A. Abdul Hamid, “Frame aggregation in wireless
networks: Techniques and issues,” IETE Technical Review, vol. 28, no. 4, pp. 336–350, 2011.

[20] K. Jamieson, B. Hull, A. Miu, and H. Balakrishnan, “Understanding the real-world performance
of carrier sense,” in Proceedings of the 2005 ACM SIGCOMM workshop on Experimental approaches to
wireless network design and analysis, 2005, pp. 52–57.

[21] A. N. Alvi, S. H. Bouk, S. H. Ahmed, and M. A. Yaqub, “Influence of backoff period in slotted
csma/ca of ieee 802.15. 4,” in International Conference on Wired/Wireless Internet Communication.
Springer, 2016, pp. 40–51.

[22] L. Chang, F. Wang, and Z. Wang, “Detection of dsss signal in non-cooperative communications,”
in 2006 International Conference on Communication Technology, 2006, pp. 1–4.

[23] hostapd configuration file. Mac frame. [Online]. Available: https://www.geeksforgeeks.org/

ieee-802-11-mac-frame/

[24] mac. mac. [Online]. Available: https://www.brainkart.com/article/IEEE-802-11 13458/

[25] D. Sharma and D. A. Karandikar, “Qos in wireless networks,” in QoS in Wireless Networks, p. 230.

[26] T. Li, T. Tang, and C. Chang, “A new backoff algorithm for ieee 802.11 distributed coordination
function,” in 2009 Sixth International Conference on Fuzzy Systems and Knowledge Discovery, vol. 3,
2009, pp. 455–459.

[27] A. Malik, J. Qadir, B. Ahmad, K.-L. A. Yau, and U. Ullah, “Qos in ieee 802.11-based wireless
networks: a contemporary review,” Journal of Network and Computer Applications, vol. 55, pp. 24–46,
2015.

[28] V. V. Kapadia, S. N. Patel, and R. H. Jhaveri, “Comparative study of hidden node problem and
solution using different techniques and protocols,” arXiv preprint arXiv:1003.4070, 2010.

[29] W. Hneiti and N. Ajlouni, “Performance enhancement of wireless local area networks,” in 2006
2nd International Conference on Information & Communication Technologies, vol. 2. IEEE, 2006, pp.
2400–2404.

[30] seemoo. Monitor mode feature. [Online]. Available: https://github.com/seemoo-lab/nexmon

[31] maui communications. monitor mode. [Online]. Available: https://www.aircrack-ng.org/doku.

php?id=airmon-ng

62 BIBLIOGRAPHY
 Appendix

[32] throughput. Sniff the mac frames. [Online]. Available: https://wireless.wiki.kernel.org/en/

developers/documentation/mac80211/ratecontrol/minstrel

[33] hostapd configuration file. fixed rate. [Online]. Available: https://w1.fi/cgit/hostap/plain/

hostapd/hostapd.conf

[34] kali. Kalilinux. [Online]. Available: https://www.offensive-security.com/kali-linux-arm-images/

[35] Wireshark-filter. Sniff the mac frames. [Online]. Available: https://dalewifisec.wordpress.com/

2014/04/29/wireshark-802-11-display-filters-2/

[36] networkdirection.net. Mtu. [Online]. Available: https://networkdirection.net/articles/

network-theory/mtu-and-mss/

[37] Wireshark. Sniff the mac frames. [Online]. Available: https://wiki.wireshark.org/CaptureSetup/

WLAN

[38] J. Sydor, “Coral: A wifi based cognitive radio development platform,” in 2010 7th International
Symposium on Wireless Communication Systems. IEEE, 2010, pp. 1022–1025.

[39] R. Ruby, S. Hanna, J. Sydor, and V. C. Leung, “Interference sensing using coral cognitive radio
platforms,” in 2011 6th International ICST Conference on Communications and Networking in China
(CHINACOM). IEEE, 2011, pp. 949–954.

BIBLIOGRAPHY 63