PERFORMANCE OPTIMIZATION OF WIRELESS MESH NETWORKS
Except where reference is made to the work of others, the work described in this thesis is my own
or was done in collaboration with my advisory committee. This thesis does not include proprietary
or classified information.
Harish Kongara
Certificate of Approval:
Shiwen Mao
Assistant Professor
Electrical and Computer Engineering
Prathima Agrawal
Samuel Ginn Distinguished Professor
Electrical and Computer Engineering
Sadasiva M.Rao
Professor
Electrical and Computer Engineering
George T. Flowers
Interim Dean, Graduate School
PERFORMANCE OPTIMIZATION OF WIRELESS MESH NETWORKS
Harish Kongara
A Thesis
Submitted to
the Graduate Faculty of
Auburn University
in Partial Fulfillment of the
Requirements for the
Degree of
Master of Science
Auburn, Alabama
May 9, 2009
PERFORMANCE OPTIMIZATION OF WIRELESS MESH NETWORKS
Harish Kongara
Permission is granted to Auburn University to make copies of this thesis at its
discretion, upon the request of individuals or institutions and at
their expense. The author reserves all publication rights.
Signature of Author
Date of Graduation
iii
VITA
Harish Kongara son of Venkateswara Swamy Kongara was born on May 23rd, 1985. He re-
ceived the Bachelor of Technology Degree in Electronics and Communications Engineering from
Acharya Nagarjuna University in 2006. He joined Masters Program in Department of Electrical
and Computer Engineering at Auburn University in August 2006. His current area of research is
focused on Wireless Mesh Networks.
iv
THESIS ABSTRACT
PERFORMANCE OPTIMIZATION OF WIRELESS MESH NETWORKS
Harish Kongara
Master of Science, May 9, 2009
(B.Tech., Acharya Nagarjuna University?India, 2006)
60 Typed Pages
Directed by Prathima Agrawal
Wireless mesh network (WMN) is a communication network consisting of radio nodes orga-
nized in a mesh topology. The components of a mesh network are mesh clients, mesh routers for
forwarding packets to mesh gateways that connect to the internet. WMNs can be integrated with
wired and wireless networks. The wireless networks can be infrastructure based networks such as
802.11 and cellular networks or mobile ad hoc networks. WMNs thus provide a low-cost platform
for offering ubiquitous broadband access ranging from local area, metropolitan area and wide area
networks. WMNs have interesting properties like self organization and self healing and offer higher
reliability. However, there are significant challenges that influence the architecture, design, and
deployment of WMNs. Clever algorithms and protocols are needed to realize the true potential of
WMNs to address performance, mobility and security issues. WMNs have applications ranging
from civilian wireless internet applications to tactical and emergency response applications
In this thesis, we focus on fairness and scalability issues in WMNs and propose solutions to
optimize network performance. Two ideas are proposed as a solution to the above problems. First,
a differentiated client algorithm is proposed as a potential solution to the scalability problem. In
this algorithm the clients are differentiated based on the amount of traffic they generate. It is shown
v
through NS-2 simulations that there is significant improvement in scalability especially when the
network is congested. Second, a gateway classification algorithm (GCA) is proposed to solve the
throughput unfairness issue. In general, WMNs are multi-hop networks. The unfairness problem
due to the fact that clients (and routers) that are nearer (smaller number of hops from) the mesh
gateways may have undue advantage over those that are farther. In this algorithm, the gateways are
classified based on their data rate handling capabilities. Extensive NS2 simulations demonstrate the
effectiveness of the above two algorithms.
vi
ACKNOWLEDGMENTS
I would express my sincere gratitude to my advisor Professor Prathima Agrawal for her guid-
ance and help during my study and research at Auburn. Thanks for leading me into the research area
of wireless mesh networks. Without her patience and support, this thesis would not be possible.
I also thank Professors Sadasiva M. Rao and Shiwen Mao for serving on my advisory com-
mittee. My thanks also go out to my colleagues in the Wireless Research Laboratory, Yogesh R
Kondareddy, Pratap Simha, Srivathsan Soundararajan and Nirmal Andrews for the discussions and
valuable suggestions on our research.
I am grateful to my parents for their consistent support and encouragement during my study
and thanks to my friends for their kind presence whenever needed.
vii
Style manual or journal used Journal of Approximation Theory (together with the style known
as ?aums?). Bibliograpy follows van Leunen?s A Handbook for Scholars.
Computer software used The document preparation package TEX (specifically LATEX) together
with the departmental style-file aums.sty.
viii
TABLE OF CONTENTS
LIST OF FIGURES xi
1 INTRODUCTION 1
1.1 Definition of Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Architecture of Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Infrastructure based Wireless Mesh Networks . . . . . . . . . . . . . . . . 2
1.2.2 Client Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.3 Hybrid Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Characteristics of Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . 3
1.3.1 Beyond ad-hoc Networking . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.2 Different types of Network Access . . . . . . . . . . . . . . . . . . . . . . 4
1.3.3 Power-Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.4 Multiple Radios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.5 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.6 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Differences between Wireless Mesh Networks and ad-hoc Networks . . . . . . . . 5
1.5 Wireless Mesh Networks Applications . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5.1 Home Based Broadband Networking . . . . . . . . . . . . . . . . . . . . 5
1.5.2 Enterprise and Community networking . . . . . . . . . . . . . . . . . . . 6
1.5.3 Other Applications of Wireless Mesh Networks . . . . . . . . . . . . . . . 7
2 STUDY OF SCALABILITY IN WIRELESS MESH NETWORKS USING NS-2 SIMULATIONS 10
2.1 Definition of Scalability in Wireless Mesh Networks . . . . . . . . . . . . . . . . 10
2.1.1 Average Network Throughput . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Characteristics of Average Network Throughput . . . . . . . . . . . . . . 12
2.2 Variation of Network Performance . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Hidden Terminal Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.2 Exposed Terminal Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.3 TCP Vs UDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.4 Error Rate vs Network Performance . . . . . . . . . . . . . . . . . . . . . 17
2.2.5 Data Rate of Clients vs Network Performance . . . . . . . . . . . . . . . . 19
2.2.6 Gateway Placement vs Network Performance . . . . . . . . . . . . . . . . 19
2.2.7 Data Rate vs Number of Clients vs Network Performance . . . . . . . . . 20
2.3 Reasons for Scalability and Fairness Issues . . . . . . . . . . . . . . . . . . . . . 22
2.4 Differentiated Clients Algorithm to Improve Scalability . . . . . . . . . . . . . . . 22
2.4.1 Algorithm for scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
ix
3 STUDY FAIRNESS IN WIRELESS MESH NETWORKS USING NS-2 SIMULATIONS 27
3.1 Definition of Fairness in Wireless Mesh Networks . . . . . . . . . . . . . . . . . . 27
3.2 Related Work in Fairness Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Analysis of the Fairness Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4 Unfairness in Different Traffic Models . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Different Traffic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 Gateway Classification Algorithm (GCA) . . . . . . . . . . . . . . . . . . . . . . 35
4 CONCLUSION 43
BIBLIOGRAPHY 44
x
LIST OF FIGURES
1.1 Infrastructure Wireless Mesh Network [1] . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Client Wireless Mesh Network [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Hybrid Wireless Mesh Network [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 WMNs for Home Networking [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Wireless Mesh Network for Transportation Systems [1] . . . . . . . . . . . . . . . 8
1.6 Wireless Mesh Network in Hospitals eliminate the need for Wired Network . . . . 8
2.1 Experimental setup demonstrating the network performance . . . . . . . . . . . . 11
2.2 Scalability in Wireless Mesh Networks . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Hidden Terminal Problem [62] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Exposed terminal problem [63] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Error Rate vs Network Performance . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.6 Data Rate vs Network Performance . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.7 Gateway Placement vs Network Performance . . . . . . . . . . . . . . . . . . . . 21
2.8 Data Rate vs Number of Clients vs Network Performance . . . . . . . . . . . . . . 21
2.9 Experimental setup for Scalability Issue . . . . . . . . . . . . . . . . . . . . . . . 24
2.10 Performance of scalability algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.11 Performance of scalability algorithm Overall Average Network Throughput . . . . 26
3.1 Idle vs Practical Fairness [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Experimental Setup Demonstrating Fairness . . . . . . . . . . . . . . . . . . . . . 29
3.3 Unfairness Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
xi
3.4 Throughput of Different Clients . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Variation of Hop Performance for Exponential Traffic . . . . . . . . . . . . . . . . 33
3.6 Variation of Hop Performance for CBR Traffic . . . . . . . . . . . . . . . . . . . . 34
3.7 Improvement in Hop Performance using Pareto Traffic Model . . . . . . . . . . . . 36
3.8 Experimental Setup for Gateway Classification Algorithm . . . . . . . . . . . . . 37
3.9 Results of Iterations without Using Gateway Classification Algorithm . . . . . . . 38
3.10 Results of Iterations with Using Gateway Classification Algorithm . . . . . . . . . 39
3.11 Throughput Variation Using Gateway Classification Algorithm . . . . . . . . . . . 40
3.12 Improvement in Throughput for Different Clients . . . . . . . . . . . . . . . . . . 41
3.13 Histogram Showing Performance of Number of Clients . . . . . . . . . . . . . . . 42
xii
CHAPTER 1
INTRODUCTION
1.1 Definition of Wireless Mesh Networks
In Wireless Mesh Networks (WMNs) each node can communicate directly with atleast one
node. Each node can also operate as a router and forward packets for other nodes that may not
be within the transmission range of their destinations. WMNs are self-organizing with the nodes
having the ability to establish an ad-hoc network and maintain the mesh connectivity [1].
1.2 Architecture of Wireless Mesh Networks
Mesh Clients and Mesh Routers are the two types of nodes in Wireless Mesh Networks. The
Mesh routers establish the backbone of the Wireless Mesh Networks while the mesh clients act as
components of the mesh environment. The mesh routers have functionalities such as gateway/bridge
functions allowing the conventional clients with wired or wireless Network Interface Cards (NICs)
to connect to the wireless mesh network[1]. Gateway or Bridge functions do not exist in wireless
Mesh Clients. Mesh clients have one wireless interface as opposed to Mesh Routers which can have
multiple interfaces. Considering all these differences the hardware platform and the software for
mesh clients is much simpler compared to mesh routers. The extension of the ad-hoc network is to
be able to connect to the Internet. The wireless mesh networks can communicate locally and as well
as to the Internet.
The architecture of WMNs is classified in to 3 types: Infrastructure WMNs, Client WMNs,
and Hybrid WMNs.
1
Figure 1.1: Infrastructure Wireless Mesh Network [1]
1.2.1 Infrastructure based Wireless Mesh Networks
In Infrastructure Wireless Mesh Networks the mesh routers form mesh infrastructure among
themselves and provides architecture for conventional clients and provides services through gate-
way/bridge functionalities. This type of architecture is less common in real world systems. The
architecture is shown in Figure 1.1 [1].
1.2.2 Client Wireless Mesh Networks
Client Wireless Mesh Networking is a synonym for ad-hoc networking. In this architecture
there are no mesh routers and clients form the network and communicate among themselves. The
clients can communicate among them locally. But they cannot connect to outside world other than
their own clients. This kind of architecture is more common in military applications. The architec-
ture is shown in Figure 1.2 [1].
1.2.3 Hybrid Wireless Mesh Networks
Hybrid Wireless Mesh Network architecture is a combination of Infrastructure and Client Wire-
less Mesh Network architecture. The clients can access the network through mesh routers or through
2
Figure 1.2: Client Wireless Mesh Network [1]
Figure 1.3: Hybrid Wireless Mesh Network [1]
other clients. This is the most common type of architecture in real world systems. The architecture
is shown in Figure 1.3[1].
1.3 Characteristics of Wireless Mesh Networks
The characteristics of WMNs are explained as follows:
3
1.3.1 Beyond ad-hoc Networking
The Wireless Mesh Networks support ad-hoc network environment with in the mesh clients
allowing them to communicate. One of the extension of the wireless mesh network is to able
to communicate to the internet. The wireless mesh networks with mesh routers connected to the
gateway makes it possible allowing the Mesh and Conventional clients to connect to the internet
[1].
1.3.2 Different types of Network Access
In WMNs, the network access is of many kinds such as peer to peer, ad-hoc, and also the
Ethernet cards supported by the mesh routers [1]. The clients communicate among themselves to
reach to a mesh router. In an ad-hoc environment there is no background architecture and all clients
communicate among themselves. Wireless Mesh Routers can be added to the wireless mesh network
environment based on the number of users in the network environment.
1.3.3 Power-Consumption
Power requirements of the wireless mesh networks is differentiated in to two types for mesh
routers the power requirement is more whereas for the mesh clients the power requirement is less
and mesh nodes can use solar energy or hydro power.
1.3.4 Multiple Radios
Mesh routers have multiple radios which separates two main types of traffic such as user traffic
and routing traffic which are handled by different radios. This improves the capacity of the network
[1].
4
1.3.5 Robustness
We expect the networks to be robust to work in any condition even though some point of the
network fails. WMNS have higher network robustness due to the availability of many mesh routers
and different paths to access the network [1].
1.3.6 Scalability
In Wireless Mesh Networks hundreds of nodes can be added at anytime to the network and the
network is completely scalable.
1.4 Differences between Wireless Mesh Networks and ad-hoc Networks
A stand alone temporary network of wireless mobile hosts over a network in which the hosts
are normally connected is called ad hoc network (Johnson and Maltz). In Wireless Mesh networks,
routing and configuration functionalities are performed by mesh routers where as end user devices
are used in ad-hoc networking to perform these functions. So, Mesh routers in Wireless Mesh
Networks result in lower power consumption and higher throughput.ad-hoc networks use single
channel for network traffic, routing and traffic configuration, this leads to reduction in overall per-
formance and it has an inherent problem to configuration and deployment for a moving user. Usage
of Multiple radios in Wireless Mesh Networks support mobile clients and also enhances network
performance.
1.5 Wireless Mesh Networks Applications
1.5.1 Home Based Broadband Networking
Now a days IEEE 802.11 WLANs are being used for home based broadband networking. Even
though these networks allow wireless data access and transfer capabilities, these networks have a
5
Figure 1.4: WMNs for Home Networking [1]
disadvantage in terms of access points, which can lead to dead zones (no service coverage). This
problem can be avoided by installing multiple access points, but this could be expensive and cumber-
some due to Ethernet wiring from access point to backhaul network modem (hub). Communication
between two user nodes in their transmission range under two different access points have to go
all the way back to the access hub, in multiple access point network. Mesh networking provides a
perfect home networking architecture. The architecture is shown in Figure 1.4 [1].
1.5.2 Enterprise and Community networking
A single office in a building or multiple rooms in the same building which are connected in
a network, represents enterprise networking. Currently data communication in these environments
are provided by IEEE 802.11. As discussed previously (Home based broad band networking) this
is also suffered by extensive usage of Ethernet cables. Even though usage of multiple modems can
increase local network capacity, but the robustness of these networks is questionable in the case of
network failure, it prohibits end user from data access or transfer. Community networks are based
6
on cable or DSL internet connection. Wireless routers, which are connected to internet modem are
being used by the end users for the internet access. This network is effective with for an individual
user, but this suffers from reduced network utilization, as all the information transfer should go
through internet with in a neighborhood. Wireless mesh networks (WMNs) can be very effective in
these scenarios, due to its flexible connections between offices and neighborhoods. Availability of
multiple paths to access internet despite of link failures can improve the network robustness.
1.5.3 Other Applications of Wireless Mesh Networks
Major advantages such as scalability and internal communication (with in the mesh clients)
have made Wireless Mesh Networks ideal for metropolitan area networking (MAN), transportation
systems, building automation, Health and Medical Systems. Mesh networks can also be imple-
mented in moving train, so that the passengers in a train can communicate among themselves and
access the internet through a gateway in the vehicle. WMNs can be useful tool in the surveillance
and health care institutions (patients data can be stored and accessed from different locations, so this
helps doctors in faster diagnosis and better services). The architecture for Transportation Systems
is shown in Figure 1.5. [1].
The Wireless Mesh Networks can also be used in various Electrical and power companies
to control the machines. In the present world the various systems in the industry are controlled
using the wired network. But the wired network is difficult to deploy and expensive, recently Wi-
Fi networks are employed but not successful due to the backend ethernet connection which results
in high cost of deployment. In these cases the wireless mesh routers can be deployed and the
information from them can be used to control the electrical machines and motors [1].
In Health and Medical systems, the data of the patient is need to processed in different rooms in
an hospital. Present day hospitals used wired connections to transmit data from one room to another
which prohibits access of the data to all rooms due to the cumbersomeness of wired networks [1]. In
7
Figure 1.5: Wireless Mesh Network for Transportation Systems [1]
Figure 1.6: Wireless Mesh Network in Hospitals eliminate the need for Wired Network
8
wireless mesh networks as shown in the Figure 1.6 the data from different rooms can be transmitted
to the data processing center using mesh routers also the data can be transmitted globally bu using
the gateway connection to the internet.
Wireless Mesh Networks can also be used in security systems where the data can be transmitted
through the mesh routers. In wired networks the entire security area need to be covered using the
wired networks. But in wireless mesh networks when an emergency occurs in a new place simply
installation of one mesh router transfers the data through the entire emergency environment [1]. This
scenario has a vital importance during emergency disasters where the firefighters and the police can
deploy one wireless mesh router to transmit the data from an unknown emergency area without the
need of installing additional wired networks[1].
The rest of the document is organized as follows. Chapter 2 introduces the scalability problem
of wireless mesh networks and analyzes the variation of performance of wireless mesh networks
with respect to data rate of clients, error rate and gateway placement.Finally a solution to the scal-
ability problem is provided at the end of the chapter.Chapter 3 introduces the unfairness issue in a
multihop wireless mesh network environment. It deals with the different traffic models that demon-
strate unfairnes among different hops.The proposed GCA is presented in this chapter which provides
a solution for the unfairness issue in wireless mesh networks. Section 4 concludes this document.
9
CHAPTER 2
STUDY OF SCALABILITY IN WIRELESS MESH NETWORKS USING NS-2 SIMULATIONS
2.1 Definition of Scalability in Wireless Mesh Networks
In a telecommunication network environment it is desired that more and more number of nodes
are added to the network anytime without the degradation of the network performance. One of
the architecture advantages of wireless mesh networks is that the network can be scalable to any
number of nodes. But the existing protocols in wireless networks are not suited for the network to
be completely scalable. SO new protocols are to be designed or existing protocols must be modified
to support the scalability of wireless mesh networks.
In WMNs, it is desired that more nodes can be added when necessary. However, with the
existing 802.11 protocols as number of nodes increase, interference increases thereby decreasing
the throughput. It is shown that the asymptotic capacity for each node decreases with the number of
nodes as O(1/n) where n is the total number of nodes in the network [2]. The network performance
depends on a number of factors, the network performance varies with the error rate, data rate of
clients, gateway placement, propagating conditions and different carrier sensing thresholds. The
experimental setup is shown in the Figure 2.1. The experimental setup consists of mesh routers,
clients and gateways in a multi hop network environment.
The scalability issue is shown in Figure 2.2. In Figure 2.2. each individual client is analyzed
for the throughput variation as more and more number of mobile clients are added to the network.
As seen from the Figure 2.2. as more and more number of nodes are added to the network the
throughput of the individual clients is reduced. This results in a decrease in the average throughput
10
Figure 2.1: Experimental setup demonstrating the network performance
of the overall network. As we can see for some the clients in the Figure 2.2., the client 6 the through-
put is dropped from 50 KBPS to 4KBPS. The results show that the existing network protocols of
802.11 are not suited for the scalability of wireless mesh networks.
2.1.1 Average Network Throughput
In this section a new definition is introduced for average network throughput. The average
network throughput is defined as the throughput per client. If the aggregate throughput of all the
clients in a network is ?x? and the number of clients in the network is ?n? then the average network
throughput is given as ?x/n?.
11
Figure 2.2: Scalability in Wireless Mesh Networks
2.1.2 Characteristics of Average Network Throughput
There are a few characteristics for average network throughput such as it decreases with in-
crease in number of clients in the network, it depends on the data rate of the clients, error rate and
gateway placement.
The Lagrange polynomial [64] can be used to estimate the variation of the network perfor-
mance based on the number of clients. The Lagrange polynomial is an interpolating polynomial of
the nth order which passes through n+1 points and is given by
L(x)=Pnj=0 yj(x)  lj(x) [64]
The Lagrange polynomial can be used to find the variation of overall network throughput. Here
we derive the Lagrange polynomial for 2 different cases. Let us derive the Lagrange polynomial for
2 different cases and observe the differences
Case 1:
12
In the case 1, let us consider a variation of mesh clients from 10 to 20. By Experimental results,
we have the network average throughput for 10 clients as 52.835Kbps. For 15 mesh clients we have
the network average throughput as 37.24Kbps, for 20 Mesh clients the network average throughput
is given as 22.03Kbps
x0 = 10:0 f(x0) = 52:835 x1 = 15:0 f(x0) = 37:24 x2 = 20:0 f(x0) = 22:03
The basis polynomials are
l0(x) = 1:0567x2  36:9845x + 317:01
l1(x) =  1:4896x2 + 44:688x 297:92
l2(x) = 0:4406x2  11:015x + 66:09
The final Lagrange polynomial is given as
L(x) = 0:0077x2  3:3115x + 85:18
If we substitute x=10 in the above equation we get the network average throughput as 52.815Kbps
which is almost same as the experimental results. If we substitute x=20 we have network average
throughput as 22.03Kbps which is same as the experimental results. From the Lagrange equation
we can find the network average through for any number of mesh clients in between 10 and 20 in
this topology. Let us verify the next case where the mesh clients are varied from 10 to 70.
Case 2: In this case if we consider ?r? mesh routers , ?g? gateways and the number of mesh
clients is varied from 10 to 70. The experimental results give us the network average through-
put as 60.283Kbps when the number of mesh clients are 10, When the number of mesh clients
are improved to 45 the network average throughput is given as 37.802Kbps, when the number of
mesh clients are improved to 70 the average network throughput is given as 17.32Kbps. Based on
these three results we can derive a polynomial which will allow us to estimate the average network
throughput for clients numbering between 10 and 70
x0 = 10:0 f(x0) = 60:28 x1 = 45:0 f(x0) = 37:802 x2 = 70:0 f(x0) = 17:32
The basis polynomials are
13
l0(x) = 0:028x2  3:3x + 90:42
l1(x) =  0:0604x2 + 4:838x 30:2416
l2(x) = 0:0115x2  0:635x + 5:196
The final Lagrange polynomial is given as
L(x) =  0:0209x2 + 0:903x + 65:3744
where x is the number of mesh clients and x varies in between 10 and 70. With the help of
the above equation we can find the average throughput of the network for any number of clients
between 10 and 70 for the above topology.
If we substitute x=10 in this Lagrange equation we get average network throughput as 72.33Kbps.
The average network throughput for 50 mesh clients is given by substituting ?x?=50 in the above
equation, we get the average network throughput as 58.27Kbps.
As seen from the Lagrange equation in this case the average network throughput derived from
Lagrange equation just gives an approximate value but not the exact value. This is due to the fact
involved as there are many number of mesh clients in the environment and we derived the equation
only with 3 experimental results. For better results the equation should be derived using more
number of experimental results. So we can say that as more number of mesh clients are involved in
the network we should use more number of experimental results to get accurate approximations.
2.2 Variation of Network Performance
In moving forward with the wireless mesh network variation with respect to different parame-
ters we would like to introduce the two most familiar problems of wireless networking the hidden
terminal and exposed terminal problems
14
Figure 2.3: Hidden Terminal Problem [62]
2.2.1 Hidden Terminal Problem
The hidden terminal problem is explained by the Figure 2.3 [62]. below the nodes A and B can
communicate to the hub, but they are hidden from eachother and as a result of this both A and B
transmit packets to the hub and a collision occurs at the hub resulting the loss of packets from both
A and B nodes.
The IEEE 802.11 uses the RTS/CTS which are request to send/ clear to send signals as a
solution to the hidden terminal problem. The sending node sends a request to send signal to the des-
tination node and the destination node send a clear to send(CTS) signal in response. Any other node
which hears the RTS/CTS should not transmit any packets during the interval of time mentioned in
the RTS/CTS. This partly solves the hidden terminal problem but they are some disadvantages with
this approach in a network environment such as wireless mesh networks in which there are very
large amount of nodes.
15
Figure 2.4: Exposed terminal problem [63]
2.2.2 Exposed Terminal Problem
The exposed terminal problem is explained by referring to the Figure 2.4 [63]. below. When
the node S1 is sending data to R1, the node S2 cannot send data to R2. The hidden and exposed
terminal problems play an important part in determining the performance of wireless networks.
2.2.3 TCP Vs UDP
TCP and UDP are the transport layer protocols used for the transmission of the packets through
the internet. TCP is used for connection-oriented transmissions and UDP is used for connectionless
transmissions. In connection-oriented transmissions a connection is established between the end
users before transmission of any data, where as in the connectionless transmissions there is no
connection established between the end users. In general connection-oriented transmissions are
called reliable transmissions. TCP is a reliable protocol compared to the UDP.
In our simulation experiments TCP is used to transfer packets from the clients to the internet.
TCP has additional advantages compared to UDP such as error correction and flow control. Flow
control determines the flow of the packets from source to destination based on the window size
of the destination node. The TCP initiates a connection through threeway hand shake. When a
transmission is through TCP it is ensured that all the packets are received in the correct order to the
destination.
16
In modern days, Wired Networks the error rate is mainly negligible. TCP assumes that the
packet loss is due to the congestion in wired networks. But in the wireless networks the error
rates are not negligible, there are lot of other factors in wireless environment resulting in the packet
loss, but TCP which is designed for wired networks basically initiates the congestion algorithms
in wireless networks when there are loss of packets. Because of the congestion algorithms the
throughput of the wireless networks is greatly reduced as a result of decrease in the window size.
2.2.4 Error Rate vs Network Performance
It is obvious that with increase in the error rate the network performance decreases. Error rate
depends on various factors and it determines the number of successful packets in total number of
packets transmitted. The packet error rate depends on the bit error rate and the collision time. In the
simulation environment used the error rate is introduced so as to calculate the effect of error rate on
network performance. In the NS2 simulation environment the error model uses a flag on the packet
to introduce errors in the transmission. There is a error rate factor in the simulation environment
which allows us to change the error rate for different simulations. In our experiment we simulated
the network for 4 different error rates randomly. For a multi-hop wireless mesh environment, the
hop performance variation with respect to the error rate is shown in Figure 2.5. As depicted from the
figure the error rate for the network affects the 3rd hop clients. As the error rate increases the loss
of packets increases and the retransmissions are increased. Due to this, the packets from the farther
hops experience more hidden terminal and exponential back-off problems, resulting in a severe loss
of throughput. The results show that the 3rd hop performance is almost zero with an increase in the
error rate. The Experimental Results are depicted in Figure 2.5.
17
Figure 2.5: Error Rate vs Network Performance
18
2.2.5 Data Rate of Clients vs Network Performance
Different mesh clients transmit at a different data rate. Based on the applications accessed by
the clients the data rate required for the successful running of application varies. So the clients which
are accessing different applications have different data rates. The overall network performance is
affected by the data rate of the individual clients. The variation of the throughput of the clients with
respect to the data rate is observed. The simulation is performed using 100 mesh clients. In addition,
the percentage of high data rate clients is varied in different simulation scenarios. Examples of some
of the applications requiring high data rate are video streaming and some examples of low data rate
applications are text messaging. In the experimental setup the percentage of high date rate clients is
taken as the measure of variation, the percentage of high data rate clients is varied in three different
scenarios and the results are plotted comparing each other. The results are shown in Figure 2.6.
In the Figure 2.6. equally distributed clients refers that half of the clients are using low data rate
applications and another half of the clients are using high data rate applications. In the second
and third scenario the high data rate clients is improved to 75 percent and 85 percent respectively.
The high data rate clients transmit data at a faster rate and hence the interference increases and the
throughput decreases, which are depicted in the experimental results.
2.2.6 Gateway Placement vs Network Performance
Also the hop performance is affected by the placement of the gateways. The number of gate-
ways and the placement of the gateways can be changed in an environment. By experimenting with
different placement of the gateways, the maximum performance is determined. In an ?n? mesh/
conventional client environment, there is a need for n/2 mesh routers and n/6 gateways and the
placement of the gateways is in such a way that the gateways always hold the center of the mesh
routers. The performance of the gateway arrangement compared to the distributed gateway is com-
pared. The results are shown in Figure 2.7. Comparing the results of the gateway placement, it
19
Figure 2.6: Data Rate vs Network Performance
shows that properly choosing the gateways will affect the performance of the overall throughput of
the network. The Results are shown in Figure 2.7.
2.2.7 Data Rate vs Number of Clients vs Network Performance
As the number of clients increases the interference increases and also as the data rate of the
clients increases the interference increases. In this case both are taken in to consideration.
The threshold throughput here is defined as the minimum average network throughput so that
each client can access the internet. If the Average network throughput is below the threshold one
or more of the network clients cannot access the internet. From the simulation results shown in the
above Figure 2.8. as the data rate of the clients is increased the threshold is reached early compared
to less data rate clients. In the above example for the datarate 1 the threshold is reached at 55 clients
that is all the 55 clients are able to access the internet. In the datarate 2 the threshold is reached for
20
Figure 2.7: Gateway Placement vs Network Performance
Figure 2.8: Data Rate vs Number of Clients vs Network Performance
21
less number of clients 38 clients that is as the data rate is increased only 38 clients are able to access
the internet. In the data rate 3 only 28 clients are able to access the internet.
2.3 Reasons for Scalability and Fairness Issues
The main reason for the fairness/scalability problems is due to the contention of the channel by
the clients. The contention of the channel is affected by many factors such as the gateway placement,
error rate, traffic model, and the type of connection (TCP or UDP). TCP needs an acknowledgement
for every packet, whereas UDP does not need an acknowledgement for every packet. The focus is
mainly given on TCP connection in this article because as the TCP connection performance is
increased in any model, the UDP performance is obviously increased. By the above results, we can
see that network throughput or the unfairness in the hops depends on many factors, including data
rate of the clients, placement of the routers/gateways, and error rate. Also, the throughput of the
clients depends on different aspects such as traffic model used and type of connection used. The
network must be scalable with the addition of clients but as the interference increases so the network
overall throughput decreases.
2.4 Differentiated Clients Algorithm to Improve Scalability
2.4.1 Algorithm for scalability
The algorithm described here is completely dynamic as more and more nodes are added to the
network at any instant of time so as to make the network scalable.It is evident that the throughput
in a wireless network is related to interference and transmission time. As the nodes are added
to the existing wireless mesh network the interference increases among the nodes. Therefore as
the number of nodes increase the throughput decreases. In this section we focus on improving
the throughput by reducing the interference. In order to increase throughput (capacity), we need
to decrease the interference by some ways. We got an idea from the sleeping node in SMAC,
22
which is for energy saving while waiting for wake-up. Our approach to decrease interference is
implemented by stopping transmission of an application and buffering for some time. It seems to
be similar to sleeping node, but different. Controlling of nodes is done in a periodic way so that
no nodes are affected in a negative way. There are many types of applications and we classified
them in to two types: high data rate applications and low data rate applications. Of course the
high data rate applications demand more capacity of the channel than low data rate ones. So we
use the low data rate applications to reduce the interference and increase the overall throughput of
network. Firstly, applications with lesser data transfer rate are identified. The less data transfer rate
applications are text messaging, browsing etc and high data rate applications are video streaming,
online TV, download of huge files etc. Secondly, transmitting nodes with lesser data rate transfer are
momentarily stopped and data is maintained in output buffer. The nodes are controlled in a periodic
way so that each low data rate node is affected in the algorithm once or twice so that its performance
is not deteriorated. The algorithm is completely dynamic and the low data rate nodes are controlled
based on different factors such as the number of clients in the network, average network throughput
and the number of transmitting nodes in the vicinity of the low data rate clients.
We used the network simulator ns2 tool for implementing our idea. The topology is given
below. We simulated a base total of 100 802.11 nodes of which 64 are mesh clients and remaining
act as mesh routers. Some of the mesh routers act as gateways through which the traffic flows to
and from the internet. In the network region the gateway mesh routers are considered to be the final
destinations. The gateways are connected to the internet by using wired connections. The wired
connections from the gateways to the internet support huge data rates and very less error rates. The
mesh routers that act as the gateways are chosen at a proper locations in the environment so that
the clients are not too far away from the gateways. The gateways are to be chosen so that as more
and more nodes are added to the network environment the clients are still able to access the internet
regardless of the interference problems. In our experiment among the 64 mesh clients 32 clients
23
Figure 2.9: Experimental setup for Scalability Issue
are chosen as high data rate applications and 32 are chosen as low data rate applications. More and
more nodes are added on and off in general case so as to make the network scalable. In general
the different clients transfer at a different data rate based on the needs of the user. The clients are
distributed uniformly in the entire network environment regardless of the high data rate or low data
rate. The experimental setup is as shown below in Figure 2.9.
The simulation procedure is as follows: N Clients N/2 high data rate and N/2 low data rate
Step1: All N active clients Step2: 3N/4 active clients (N/4 low data rate active) Step3: N/2 active
clients (N/2 low data rate inactive)
The above procedure is explained as follows in the step1 all the clients are active that is there
is no algorithm implemented in the step1. In general the network performance is determined based
on the actual environment. In this case all the 64 mesh clients are active, none of the low data rate
mesh clients are controlled.
24
In the step2 N/4 low data rate clients are controlled based on the vicinity of the client, average
network throughput and the number of high data rate clients that are actually transmitting data. In
this case 8 low data rate clients are controlled in a period way out of the 32 low data rate clients.
The 8 low data rate clients are controlled in such a way that none of the performance is deteriorated.
The application is buffered and made sure the packet reaches before the end of the timer in a TCP
connection.
In the Step3 N/2 low data rate clients are controlled based on the algorithm procedure so as to
improve the performance of the overall network. In this case 16 low data rate clients are controlled.
In the above cases more and more nodes are added to the network environment at any instant
making the network scalable. The number ?N? is varied in real world environment from time to
time.
We compared the throughput in the three cases and we got the results as follows. We ana-
lyzed our idea for mobile nodes and stationary nodes. The simulation results are as shown in the
Figure 2.10.
As seen in the experimental results the throughput of different clients is varied in all the above 3
steps. In this case the overall average network throughput is also increased in the 3 steps.The overall
average network throughput variation in all the above three steps is as shown in the figure 2.11.
25
Figure 2.10: Performance of scalability algorithm
Figure 2.11: Performance of scalability algorithm Overall Average Network Throughput
26
CHAPTER 3
STUDY FAIRNESS IN WIRELESS MESH NETWORKS USING NS-2 SIMULATIONS
3.1 Definition of Fairness in Wireless Mesh Networks
In wireless mesh networks the nodes farther away from the gateway may be starved by nodes
closer to the gateway[3]. This is due to the fact that the clients that are farther hops away exhibit
frequent exponential back off. This leads to a higher probability of collision and loss, and a corre-
sponding throughput decrease. This leads to unfairness among different clients in different hops in
sharing of the throughput of the channel.
3.2 Related Work in Fairness Issue
There are various methods proposed to modify the back-off algorithm and contention windows
for improving fairness which can be found in [59] [61]. One more approach NAV-blocking in
which each Access Point transfers a beacon which does not allow the mobile stations to transmit
during that interval [7]. The Forced Handoff technique forces the mobile stations between different
Base Stations to switch between different base stations to achieve a fair throughput [7]. We used a
Gateway Classification Algorithm in which the gateways are differentiated to handle different types
of traffic.
3.3 Analysis of the Fairness Issue
In cellular networks, the network area is divided in to cells and each mobile user is only one hop
away from its base station [11]. The other mobile users need not transmit the traffic of other mobile
27
Figure 3.1: Idle vs Practical Fairness [2]
users to the base station. But in a wireless mesh network environment a mobile client user node
has to transmit relayed traffic and also its own traffic. In a wireless mesh network the mesh clients
contend among one another for the destination node or mesh router which transmits the traffic to
the gateway and also between its traffic and the relayed traffic [2]. The contention environment is
different in wireless LANs in infrastructure mode and cellular networks because the user nodes are
only at one-hop distance from the access point and base station respectively[2].
The fairness problem is clearly explained by the Figure 3.1[2] there are two nodes,node 1 and
node 2 which are offered a load of G. The destination node for the two nodes is the gateway node
(GW) as shown in the figure. In the ideal case the channel is shared equally among the 2 nodes as
the offered load (G) increases as shown in the Figure 3.1. But in practical case as the mesh clients
have to contend between the relay traffic and its own traffic the node that is closer to the gateway
node 1starves the node far away from the gateway The result in Figure 3.1 is obtained under the
fact that the traffic to be forwarded by node 1 originating from the node 2 is queued together with
the traffic from its own node 1[2]. So the correlation between different mesh clients fail to provide
end-to-end fairness in multihop wireless mesh networks.
In a Wireless Mesh Network, there are several gateways. The gateways along with the mesh
routers form a multi-hop environment for the mesh and conventional clients. We consider the topol-
ogy as shown in the following Figure 3.2.
28
Figure 3.2: Experimental Setup Demonstrating Fairness
The topology consists of a gateway and mesh routers which form the multi-hop environment.
The clients shown in the Figure 3.2. can communicate to the gateway through the mesh routers.
The clients that are not within the reach of mesh router communicate with different mesh clients
to forward the packets to the mesh router. Each client operates not only as a host but also a router
forwarding packets on behalf of other clients. The clients in each hop are traveling at different
speeds, ranging from pedestrian to vehicular speeds. Also, the clients are communicating to the
gateway connected to the Internet at different data rates. Each client can access different applications
such as browsing, text message, online TV, download of huge files which require different data rates
for different clients. We used network simulator (NS-2) for simulation. The connection used here
is a TCP connection in which the packets are acknowledged. The clients here are of different data
rates to match to the traffic of the different users such as video streaming, text messaging. The
throughput of different clients belonging to different hops is measured. The communication used
29
Figure 3.3: Unfairness Issue
here is a TCP communication which requires all the packets to be acknowledged back to the client.
The transmission time for each successful packet is the total time in between release of the packet
from the client and the time in which the acknowledgement is received. The throughput variation
of the different hop clients is shown in Figure 3.3.
The Figure 3.3. shows that the throughput varies among different hops. The farther away hops
experience frequent exponential back-off than the nearby hops as stated in [3]. The farther away
hops from the gateway are starved due to the near hops from the gateway. The throughput variation
of the first and the third hops differs a lot as shown in the Figure 3.3.
As the data rate of clients increases the far away hops are starved. If the data rate of the clients
is very less than the clients up to 5th or 6th hops can access the gateway. The Lagrange polynomial
30
Figure 3.4: Throughput of Different Clients
as described in chapter 2 can also be used here to find the equation of the hop throughput for general
nth hop.
A simulation environment with 100 mesh clients traveling at different speeds and 30 mesh
routers in which three routers (10 percent of the total routers) act as gateways are randomly deployed
and the performance of the clients is measured. It is observed that different clients have different
throughputs which demonstrates the unfairness issue. As seen from the Figure 3.4. eventhough
all the clients are in the same network area they are not able to share the resources fairly. The
throughput varies depending on the hop of the client as illustrated in Figure 3.4.
31
3.4 Unfairness in Different Traffic Models
3.4.1 Different Traffic Models
There are different types of traffic models used in the real network environment such as Expo-
nential, Pareto and CBR.
The Exponential ON/OFF traffic generator in a simulation environment has different parame-
ters such as packetsize,bursttime, idletime and rate. The burst time is the average ?on? time for the
traffic generator, the idle time is the average ?off? time for the generator and rate is the sending rate
of the packets during the ?on? time [48].
In ns-2 simulation environment the Exponential On/Off generator can be made to work as a
Poisson generator by changing the variable bursttime to 0 and by changing the variable rate to a
very large value. The C++ code ensures that even if the burst time is set to zero, at least one packet
is sent [48]. And also the next interarrival time is the sum of the assumed packet transmission time
and the random variate corresponding to idle time[48]. To make the first term in the sum very small,
make the burst rate very large so that the transmission time is almost zero compared to the typical
idle times [48].
The CBR traffic generator has different parameters such as packetsize, rate and random. The
rate is the constant rate at which the packets are sent, random is the flag indicating whether or not
to introduce random ?noise? in the scheduled departure times [48].
The fairness issue is demonstrated for different types of user traffic models. The exponential
and CBR traffic models are used as the user traffic. The following results show the Unfairness for
different traffic models. In the first case Figure 3.5 all the clients use Exponential traffic and in the
second case Figure 3.6 all the clients use CBR traffic. As seen from the results both the exponential
and CBR traffic models have unfairness issue in the hops. If there is some factor to control the
32
Figure 3.5: Variation of Hop Performance for Exponential Traffic
network traffic, we could have avoided the fairness to some extent. There is no factor as to control
the unfairness in these type of traffic models.
Earlier in this section we performed the experiment on Exponential and CBR traffic models.
Now we have focused our attention on the Pareto Traffic Model which allow us to modify some
factors to improve the fairness of the far way hops. The Pareto On/Off Traffic Generator is a traffic
generator embodied in the OTcl class Application/Traffic/Pareto. It generates traffic according to
a Pareto On/Off distribution. Packets are sent at a fixed rate during on periods, and no packets
are sent during off periods. Both on and off periods are taken from a Pareto distribution with
constant size packets. These sources can be used to generate aggregate traffic that exhibits long
range dependency. There are various input parameters in the pareto traffic model such as burst time,
33
Figure 3.6: Variation of Hop Performance for CBR Traffic
34
idle time, rate, packet size and shape. In each On/Off round the next burst time and next idle time are
computed based on the pareto shape parameter. The next burst time refers to the number of packets
to be transmitted in the next burst period. Based on the characteristics of the pareto traffic model
the numbers of packets send during the on-time of an application can be varied.In a Wireless Mesh
Network as we know the farther away hops are starved by the hops closer to the gateway. Using
the characteristics of pareto traffic model the farther away hops are made to send more number of
packets during the on period than the hops that are closer to the gateway by varying the pareto shape
parameter. This improves the performance of the far away hops as compared to without using this
algorithm. Now here the Pareto shape parameters are setup in such a way that the clients that are
father hops away from a gateway are made to send more number of packets during the on time as
compared to the clients that are nearer to the gateway. This algorithm is applied to observe the
improvement in the performance of the far away hops. The experimental results are as shown in
Figure 3.7.
3.5 Gateway Classification Algorithm (GCA)
In Chapter 2 the network performance is observed by properly placing the gateways in a net-
work environment, the gateways are not classified into different types. The results are obtained by
changing the position of the gateways. In the results, we can observe a change in the throughput
of the overall network. This gave us an idea that the gateways can be modified in order to vary the
network performance. In this section, we classify the gateways into different types. In a wireless
mesh networks, there are several gateways. We classify the gateways into two types: high data rate
gateways and low data rate gateways. The high data rate gateways are responsible for handling the
traffic of the high data rate clients. The low data rate gateways are responsible for handling the low
data rate clients. By classifying the gateways into these two types, we can improve the performance
in terms of fairness. The traffic from the clients is routed to the appropraite gateway based on the
35
Figure 3.7: Improvement in Hop Performance using Pareto Traffic Model
36
Figure 3.8: Experimental Setup for Gateway Classification Algorithm
traffic of the clients. As shown in the experimental setup, the gateways are classified and the traffic
is routed to an appropriate gateway. A number of experiments have been performed to study the
fairness issue using the proposed GCA. As stated earlier, the ns-2 simulator is used to perform the
experiments. The experimental setup is depicted in Figure 3.8.
The gateways as indicated in the figure are classified in to two types. The improvement of the
overall network performance is shown in Figure 3.11.
The Experimental results during the different iterations are as shown in the figure. The Exper-
iment is iterated for a total runs of 100 and the average of all the results is shown in the Figure 3.9.
The results of some of the interations are as shown in the Figure 3.10.
As shown in the experimental setup, the different clients are routed to different gateways.
The clients are now neither far away nor near as in the normal architecture as the gateways are
distributed throughout the entire network. The traffic is routed in such a way that all the low data
rate traffic is routed to one of the type of gateways and all high data rate traffic is routed to other
37
Figure 3.9: Results of Iterations without Using Gateway Classification Algorithm
38
Figure 3.10: Results of Iterations with Using Gateway Classification Algorithm
39
Figure 3.11: Throughput Variation Using Gateway Classification Algorithm
type of gateways. The clients do not experience the exponential back-off frequently as in the normal
architecture. Because of this, the throughput of the overall network is distributed in a reasonable
manner. As depicted in the Figure 3.12., the variation of the throughput of the clients in the two
different cases and the overall performance distributed in a fair manner. The variation of throughput
of different clients is shown in Figure 3.12.
In the above Figure 3.12. the variation of throughput of each individual client is observed after
the implementation of the algorithm. It can be obviously seen that the performance of each client is
improved based on the gateway classification algorithm.
The histogram in Figure 3.13. shows the number of clients at different throughput ranges.
We can observe the increase in the number of clients in the range of 2nd hop throughput. As
shown in the Figure 3.13., the throughput variation is huge. Without GCA the number of clients
in the throughput range (0-10) are 83. After the application of the proposed GCA, the number of
40
Figure 3.12: Improvement in Throughput for Different Clients
clients are only 23. This shows that more clients that are in the lower data rate range are shifted
to the higher data rate range, which supports the fact that more channel share is provided to those
clients. Also, the number of clients in the throughput range of (10-20) are increased from 13 to 68.
With the proposed GCA, the throughput of the 3rd hop clients is almost increased three times. In
our simulation results the network performance of the third hop clients is improved by a factor of
6.5.
With the implementation of the Gateway Classification Algorithm the throughput of the far
away hops can be significantly improved and the overall network performance can be fair. The
algorithm proposed can be implemented without the need for additional resources and by changing
the protocols involved in the network environment.
41
Figure 3.13: Histogram Showing Performance of Number of Clients
42
CHAPTER 4
CONCLUSION
Wireless mesh networks have many advantages, and also there are other issues which need
more investigation. Other issues associated with the mesh networks are scalability, mobility, unfair-
ness and security. More investigation is needed in these areas. In the first chapter of this thesis, we
presented the characteristics, applications and problems associated with wireless mesh networks.
In the second chapter we analyzed the variation of the performance of wireless mesh networks
with respect to different parameters such as error rate, gateway placement, data rate of clients. In
this chapter we also demonstrated the scalability issue and proposed a differentiated clients algo-
rithm to solve the scalability problem. The algorithm proposed is a dynamic algorithm and the
performance improves as more and more mesh clients are added to the network environment. The
algorithm is verified using the NS-2 simulation environment and the results obtained show a signif-
icant improvement in the performance.
In the third chapter the unfairness in multi-hop wireless mesh networks is introduced and the
performance of different traffic models such as the exponential, Constant Bit Rate and the Pareto
traffic models on different hops is observed. Finally a Gateway classification algorithm is proposed
which differentiates the gateways in the environment and the traffic is routed to the appropriate
gateway. The simulation is performed using ns-2 simulator and the results obtained show a signifi-
cant improvement in the performance of the far way hop clients. In the future the improvement in
throughput performance when both, differentiated client and differentiated gateway algorithms are
employed will be studied and heterogenous clients will be deployed to emulate realistic scenarios.
43
BIBLIOGRAPHY
[1] I.F. Akyildiz and X. Wang, ?A Survey on Wireless Mesh Networks,? IEEE Communications
Magazine, September 2005.
[2] J. Jun, ?The nominal capacity of wireless mesh networks,? Wireless Communications, IEEE
Volume 10, Issue 5, September 2005.
[3] V. Gambiroza, B. Sadeghi and E. Knightly, ?End-to-End Performance and Fairness in Multi-
hop Wireless Backhaul Networks,? Proceedings of ACM MobiCom 2004, Philadelphia, PA,
September 2004.
[4] C.W. Hsu, C.Y. Wang and T.C. Hou, ?Providing End-to-End Fairness in Wireless Mesh Net-
works with Chain Topologies,? Parallel and Distributed Computing and Systems, November
2006.
[5] N. Bisnik, and A. Abouzeid, ?Delay and throughput in random access wireless mesh net-
works,? Proceedings of IEEE International Conference on Communications ICC?06, June
2006.
[6] D. Zhao, ?Inter-AP coordination for fair throughput in infrastructure-based IEEE 802.11 mesh
networks,? International Conference on Wireless Communications and Mobile Communica-
tions, Vancouver, Canada, July 2006.
[7] D. Zhao, ?Achieving fair throughput in infrastructure-based IEEE 802.11 mesh networks,?
IEEE GLOBECOM?06, December 2006.
[8] R.G. Cheng, C.Y. Wang and L.H. Liao, ?Ripple: A Distributed Medium Access Protocol for
Multi-hop Wireless Mesh Networks,? Vehicular Technology Conference, September 2006.
[9] B. Braunstein, T. Trimble, R. Mishra, B.S. Manoj and R. Rao, ?On The Traffic Behavior
of Distributed Wireless Mesh Networks,? International Symposium on a World of Wireless,
Mobile and Multimedia Networks, June 2006.
[10] R. Cheng, C. Wang, L. Liao, and J. Yang, ?Ripple: A wireless token-passing protocol for
multi-hop wireless mesh networks,? Communications Letters, IEEE, September 2006.
[11] N.B. Salem and J.P. Hubaux, ?Securing wireless mesh networks,? Communications Letters,
IEEE, September 2006.
[12] B. Aoun, R. Boutaba and G. Kenward, ?Analysis of Capacity Improvements in Multi-Radio
Wireless Mesh Networks,? Vehicular Technology Conference, September 2006.
44
[13] R.K. Lam, D. Chiu and C.S. Lui, ?On the Access Pricing Issues of Wireless Mesh Networks,?
International Conference on Distributed Computing Systems, July 2006.
[14] T. Wu, Y. Xue and Y. Cui, ?Preserving Traffic Privacy in Wireless Mesh Networks,? Interna-
tional Symposium on a World of Wireless, Mobile and Multimedia Networks, June 2006.
[15] A. Acharya, A. Misra and S. Bansal, ?Design and analysis of a cooperative medium access
scheme for wireless mesh networks,? First International Conference on Broadband Networks,
2004.
[16] R. Luo, D. Belis, R.M. Edwards and G.A. Manson, ?simulation design for link connection-
oriented wireless mesh networks,? 4th International Workshop on Mobile and Wireless Com-
munications Network, 2002.
[17] P. Beckman, S. Verma and R. Rao, ?Use of mobile mesh networks for inter-vehicular commu-
nication,? Vehicular Technology Conference, 2003.
[18] A. Raniwala and T. Chiueh, ?Evaluation of a wireless enterprise backbone network architec-
ture,? 12th Annual IEEE Symposium on High Performance Interconnects, 2004.
[19] S. Naghian, ?Mesh vs. point-to-multipoint topology: a coverage and spectrum efficiency com-
parison,? 15th IEEE International Symposium on Personal, Indoor and Mobile Radio Com-
munications, 2004.
[20] L. Iannone, R. Khalili, K. Salamatian and S. Fdida,?Cross-layer routing in wireless mesh
networks,? International Symposium on Wireless Communication Systems, 2004.
[21] A. Adya, P. Bahl, J. Padhye, A. Wolman, L. Zhou, ?A multi-radio unification protocol for IEEE
802.11 wireless networks,? First International Conference on Broadband Networks, 2004.
[22] F. Tasaki, H. Tamurat, M. Sengokut and S. Shinodas, ?A new channel assignment strategy
towards the wireless mesh networks,? 5th International Symposium on Multi-Dimensional
Mobile Communications Proceedings, 2004.
[23] P.H. Hsiao and H.T. Kung, ?Layout design for multiple collocated wireless mesh networks,?
Vehicular Technology Conference, 2004.
[24] M.K. Marina and S.R. Das, ?A topology control approach for utilizing multiple channels in
multi-radio wireless mesh networks,? 2nd International Conference on Broadband Networks,
2005.
[25] V. Navda, A. Kashyap and S.R. Das, ?Design and evaluation of iMesh: an infrastructure-mode
wireless mesh network,? IEEE International Symposium on a World of Wireless Mobile and
Multimedia Networks, 2005.
[26] A.K. Das, H.M. K. Alazemi, R. Vijayakumar and S. Roy, ?Optimization models for fixed chan-
nel assignment in wireless mesh networks with multiple radios,? 2005 Second Annual IEEE
Communications Society Conference on Sensor and Ad Hoc Communications and Networks,
2005.
45
[27] B. Ji, ?Asynchronous wireless collision detection with acknowledgement for wireless mesh
networks,? Vehicular Technology Conference, 2005.
[28] L. Liu and G. Feng, ?Research on Multi-constrained QoS Routing Scheme Using Mean Field
Annealing,? Sixth International Conference on Parallel and Distributed Computing, Applica-
tions and Technologies, 2005.
[29] C. Zhu, M.J. Lee and T. Saadawi, ?On the route discovery latency of wireless mesh networks,?
Consumer Communications and Networking Conference, 2005.
[30] K. Balaji, N. Hegde, B. Venkata Ramana, B. S. Manoj, and C. Siva Ram Murthy,?Performance
evaluation of a hybrid wireless network architecture for rural communication,? IEEE Interna-
tional Conference on Personal Wireless Communications, 2005.
[31] F.H.P. Fitzek, T.K. Madsen, R. Prasad and M. Katz, ?Cooperative IP header compression for
parallel channels in wireless meshed networks,? IEEE International Conference on Communi-
cations, 2005.
[32] P.M. Ruiz and A.F. G. Skarmeta, ?Approximating optimal multicast trees in wireless multihop
networks,? IEEE Symposium on Computers and Communications, 2005.
[33] R. Liscano, E.F. Sadok and E.M. Petriu , ?Mobile wireless RSA overlay network as critical
infrastructure for national security,? IEEE International Workshop on Measurement Systems
for Homeland Security, Contraband Detection and Personal Safety Workshop, 2005.
[34] I.F. Akyildiz, ?Key technologies for wireless networking in the next decade,? International
Conference on Pervasive Services, 2005.
[35] L. Fu, Z. Cao and P. Fan, ?Spatial reuse in IEEE 802.16 based wireless mesh networks,? IEEE
International Symposium on Communications and Information Technology, 2005.
[36] S. Thuel and R. Cruz, ?The next big bang in wireless and mobile communication,? 14th Inter-
national Conference on Computer Communications and Networks, 2005.
[37] K.N. Ting, Y.F. Ko and M.L. Sim, ?Voice performance study on single radio multi-hop IEEE
802.11b systems with chain topology,? 13th IEEE International Conference on Networks
jointly held with IEEE 7th Malaysia International Conference on Communication, 2005.
[38] C.E. Seo, E.J. Leonardo, P. Cardieri, M.D. Yacoub, D.M. Gallego and A.A.M. de Medeiros,
?Performance of IEEE 802.11 in wireless mesh networks,? International Conference on Mi-
crowave and Optoelectronics,2005.
[39] W. Allen, A. Martin, A. Rangarajan, ?Designing and deploying a rural ad-hoc community
mesh network testbed,? The IEEE Conference on Local Computer Networks, 2005.
[40] Y. Ma, ?Improving wireless link delivery ratio classification with packet SNR,? International
Conference on Electro information Technology, 2005.
[41] G. Goth, ?Groups hope to avoid mesh standard mess,? Distributed Systems online, IEEE 2005.
46
[42] B. Cheng, S. Kalyanaraman and M. Klein, ?A geography-aware scalable community wireless
network test bed,? First International Conference on Testbeds and Research Infrastructures for
the Development of Networks and Communities, 2005.
[43] I. Aydin, C. Jaikaeo, and C. Shen ,?Quorum-based match-making services for wireless mesh
networks,? IEEE International Conference on Wireless and Mobile Computing, Networking
and Communications, 2005.
[44] J.A. Stine, ?Exploiting smart antennas in wireless mesh networks using contention access,?
Wireless Communications, IEEE, 2006.
[45] C. Ma, Z. Zhang and Y Yang, ?Battery-aware router scheduling in wireless mesh networks,?
20th International Parallel and Distributed Processing Symposium, 2006.
[46] Jane- Hwa. Huang, Li-Chun. Wang and Chung-Ju. Chang, ?Coverage and capacity of a wire-
less mesh network,? International Conference on Wireless Networks, Communications and
Mobile Computing, 2005.
[47] T. Tsai and J. Chen, ?IEEE 802.11 MAC protocol over wireless mesh networks: problems
and perspectives,? 19th International Conference on Advanced Information Networking and
Applications, 2005.
[48] NS-2 manual documentation, ?http://www.isi.edu/nsnam/ns/doc/index.html?.
[49] M. Kodialam and T. Nandagopal, ?Characterizing Achievable Rates in Multi-Hop Wireless
Mesh Networks With Orthogonal Channels,? Transactions on Networking, IEEE, 2005.
[50] X. Tao, T. Kunz and D. Falconer,?Traffic balancing in wireless MESH networks,? International
Conference on Wireless Networks, Communications and Mobile Computing, 2005.
[51] L. Liu and G. Feng, ?Mean field network based QoS routing scheme in wireless mesh net-
works,? International Conference on Wireless Communications, Networking and Mobile Com-
puting, 2005.
[52] Z. Lei, Y. ShouBao, S. Weifeng and W. DaPeng, ?Markov-Based Analytical Model for TCP
Unfairness over Wireless Mesh Networks,? First International Multi-Symposiums on Com-
puter and Computational Sciences, 2006.
[53] B. Wehbi, W. Mallouli and A. Cavalli, ?Light Client Management Protocol for Wireless Mesh
Networks,? International Conference on Mobile Data Management, 2006.
[54] H. Viswanathan and S. Mukherjee, ?Throughput-range tradeoff of wireless mesh backhaul
networks,? IEEE Journal on Selected areas in Communications, 2006.
[55] N. Funabiki, S. Yoshida, S. Tajima and T. Higashino, ?An Internet Gateway Access-Point
Selection Problem for Wireless Infrastructure Mesh Networks,? 7th International Conference
on Mobile Data Management, 2006.
[56] D.M. Shrestha and Y.B. Ko, ?On Construction of the Virtual backbone in Wireless Mesh Net-
works,? The 8th International Conference Advanced Communication Technology, 2006.
47
[57] M.J. Lee, J. Zheng, Y.B. Ko and D.M. Shrestha, ?Emerging standards for wireless mesh tech-
nology,? Wireless Communications, IEEE, 2006.
[58] T. Tsai, H. Tseng and A. Pang, ?A new MAC protocol for Wi-Fi mesh networks,? 20th Inter-
national Conference on Advanced Information Networking and Applications, 2006.
[59] R.E. Ludwig, ?Eliminating inefficient cross-layer interactions in wireless networking,? Ph.D.
dissertation, Aachen University of Technology, Germany, 2000.
[60] V. Bharghavan, A. Demers, S. Shenker, and L. Zhang, ?MACAW: a media access protocol for
wireless LANs, in Proc. of ACM SIGCOMM94, 1994.
[61] J. Liu and S. Singh, ?ATCP:TCP for mobile ad hoc networks,? IEEE Journal on selected areas
in communications, vol. 19, no. 7,pp, 1300-1315, July 2001.
[62] www.wikipedia.com, ?http://upload.wikimedia.org/wikipedia/en/thumb/b/b1/Hidden-
node.svg/524px-Hidden-node.svg.png?.
[63] www.wikipedia.com, ?Exposed Terminal Problem?.
[64] A. Quarteroni, R. Sacco, and F. Saleri, ?Numerical Mathematics (Texts in Applied Mathemat-
ics) (Hardcover)?, November, 2006.
48