Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
Ants Inspired Traffic Classification in Mobile Ad Hoc Networks
by
Engr. Babar Nazir
In Partial Fulfillment of the Requirements for the Degree
of Doctor of Philosophy
Department of Electrical Engineering
University of Engineering and Technology Peshawar,
PAKISTAN
June, 2011
ii
© Copyright by University of Engineering and Technology Peshawar,
Pakistan 2011
iii
Dedicated to my Mother
All that I am or hope to be, I owe to my angel mother.
iv
ACKNOWLEDGEMENTS
All praise to Allah, the Almighty, the most merciful and compassionate, who granted
me perseverance to accomplish this research work.
I utter my profound thanks to my supervisor Prof. Dr. Azzam-ul-Asar for accepting
me as his student. His motivation guided me to this achievement and kept my morale
high with his valuable suggestions. He was readily available for consultations devoid
of that I would have never been capable to complete this work.
I am also thankful to my REC members, Prof. Dr. Iftikhar Ahmed, Prof. Dr. M. A.
Irfan Mufti, Dr. Mehmood Ashraf and last but not the least Dr. Tariq Mehmood Raja,
all of whom provided me valuable suggestions during my all the REC meetings held
in connection with my this research work.
I must also thank Prof. Dr. Waqar Shah Director, Post Graduate Studies for his
inspiring attitude and guidance.
My thanks also go to Dr. David Harle of University of Strathclyde and Dr. Hasib Zafar
for their support during my six months stay at University of Strathclyde.
I am also thankful to my foreign evaluators namely Dr. Mudasser Fraz Wyen and Dr.
Shakil Akhtar for reviewing this thesis and their expert views for guidance.
I would also like to thank my parents, teachers and friends and all those who helped
me and supported me during the course of my studies.
v
Abstract
Mobile ad hoc network is new paradigm in communication networks. Unlike wired or
internet, these are wireless networks consisting of mobile nodes which are rapidly
deployable and self organizing in nature. Communication between network nodes
takes place through scarce wireless channels. Each node is capable of relaying the
traffic of other nodes, so called multi hop network.
The rising popularity and commercial usage of MANETs demands support for Quality
of Service in such networks to work effectively. Supporting QoS in such ad hoc
environments has been broadly accepted as a tricky problem due to particular
uniqueness of these networks such as dynamics of the environment, node mobility,
and improbability of availability of resources.
Active research is underway in the area of providing quality of service in mobile ad
hoc networks but work is still considered in its nascent stage, which can broadly be
categorized in two main streams. One deals with proposing extensions in already
proposed best effort routing protocols, while other is dedicated towards developing
purpose built QoS aware routing protocols based on conventional routing approaches.
Both of these strategies are unable to offer a mature solution for dynamics of MANET
environment.
Recent research in swarm intelligence has revealed matching properties between the
routing requirements of communication networks and certain tasks that exist in social
vi
insect colonies. Research community has proposed promising solutions for routing in
dynamic ad hoc networks on the basis of these swarm intelligence principles.
In this thesis, in the first stage, simulation has been carried out for performance
evaluation of two leading MANET routing protocols to emphasize the ineffectiveness
of conventional routing techniques. In the second stage, a novel QoS provisioning
scheme called Ants Inspired Traffic Classification in Mobile Ad Hoc Networks has
been developed which is based on Swarm intelligence obtained through stigmergy
process found in nature like ant colony. Most of the conventional routing approaches
heavily rely on feedback from other layers while this approach is layer independent in
nature.
Service classes are made available through the proposed solution in order to serve
according to priority of the traffic. The solution uses an adaptive approach giving
alternate QoS supporting paths, which are instantly made available in case of degraded
performance that often happens in dynamic environment of MANET. Redundant
routes often appear in MANET’s dynamic environment and remain present in routing
tables thus increasing overheads. The proposed solution addresses this issue through
evaporation property of Pheromone in Swarm Intelligence.
In our case, the most important parameter of QoS is considered in terms of Bandwidth.
Our SI based protocol defines traffic paths in the network of various classes belonging
to different range of bandwidths.
vii
Table of Contents Table of Contents ..................................................................................................... vii
List of Figures .......................................................................................................... xii
List of Tables........................................................................................................... xvi
CHAPTER 1: INTRODUCTION ............................................................................ 1
1.1 General ....................................................................................................... 1
1.2 Problem Statement ...................................................................................... 1
1.3 Objectives and Scope .................................................................................. 3
1.4 Original Contributions ................................................................................. 4
1.5 Organization of Thesis ................................................................................ 5
CHAPTER 2: TECHNICAL BACKGROUND ....................................................... 7
2.1 Mobile Ad Hoc Networks ........................................................................... 7
2.2 Mobile Ad Hoc Networks Protocol Layer Stack .......................................... 9
2.3 Mobile Ad Hoc Networks Applications ..................................................... 11
2.3.1 Applications in Natural Disasters........................................................... 11
2.3.2 Battlefield Scenarios.............................................................................. 12
2.3.3 Enterprise Networks .............................................................................. 13
2.4 MANET Characteristics ............................................................................ 14
2.4.1 Scarce radio bandwidth ............................................................................ 14
2.4.2 Distributive and without any infrastructure ............................................... 15
2.4.3 Dynamic topology and mobility ............................................................... 15
2.4.4 Little contact duration among nodes ......................................................... 15
2.4.5 Economical solution ................................................................................. 16
viii
2.4.6 Beyond LOS connectivity ........................................................................ 16
2.4.7 Numerous nodes and huge network .......................................................... 16
2.4.8 Bit error rate ............................................................................................. 16
2.5 MANET Challenges .................................................................................. 17
2.5.1 Dynamic Topology ................................................................................... 17
2.5.2 Asymmetrical Links ................................................................................. 17
2.5.3 Frequent link breakage ............................................................................. 18
2.5.4 Hidden terminal and exposed terminal problem ........................................ 18
2.5.5 Security issues .......................................................................................... 20
2.5.6 Distributed operations .............................................................................. 20
2.6 Quality of Service ..................................................................................... 20
2.6.1 Challenges for providing QoS: ................................................................. 23
2.7 Conclusion ...................................................................................................... 25
CHAPTER 3: ROUTING RELATED WORK ...................................................... 26
3.1 General ........................................................................................................... 26
3.2 MANET Routing Protocols ....................................................................... 30
3.2.1 AODV (Ad Hoc On Demand Distance Vector)...................................... 30
3.2.2 DSDV (Destination Sequenced Distance vector) ................................... 33
3.2.3 DSR (Dynamic Source Routing) ............................................................ 35
3.2.4 OLSR (Optimized Link State Routing) .................................................. 38
3.2.5 ZRP (Zone Routing Protocol) ................................................................ 39
3.3 QoS Aware Routing .................................................................................. 40
3.4 Conclusion ................................................................................................ 44
ix
CHAPTER 4: DSDV AND DSR SIMULATION.................................................. 46
4.1 General ........................................................................................................... 46
4.2 Simulation Environment ................................................................................. 47
4.3 Simulation Overview................................................................................. 48
4.3.1 Pre-Simulation Phase ............................................................................ 50
4.3.2 Ns-2 Execution Phase ............................................................................ 50
4.3.3 Post-Simulation phase ........................................................................... 50
4.4 Simulation Setup ....................................................................................... 51
4.4.1 DSDV Simulation ................................................................................. 51
4.4.2 DSR Simulation .................................................................................... 56
4.5 DSR and DSDV Result Analysis ..................................................................... 60
4.6 Measures of Interest .................................................................................. 60
4.6.1 Data throughput ..................................................................................... 61
4.6.2 Routing Overhead ................................................................................. 61
4.6.3 Packet Drop Ratio ................................................................................. 61
4.7 Mobility Analysis ...................................................................................... 61
4.7.1 Throughput Vs Mobility ........................................................................ 62
4.7.2 Protocol Overhead Vs Mobility ............................................................. 64
4.7.3 Packet Drop Ratio Vs Mobility .............................................................. 65
4.8 Load Analysis ........................................................................................... 66
4.8.1 Throughput Vs Load ............................................................................. 67
4.8.2 Protocol Overhead Vs Load ................................................................... 68
4.8.3 Packet Drop Ratio Vs Load ................................................................... 69
x
4.9 Conclusion ...................................................................................................... 70
CHAPTER 5: SWARM INTELLIGENCE ............................................................ 73
5.1 General ........................................................................................................... 73
5.2 Advantages of Swarm Intelligence .................................................................. 74
5.3 SI based Routing ............................................................................................. 76
5.4 Conclusion ...................................................................................................... 82
CHAPTER 6: PROPOSED SOLUTION ............................................................... 84
6.1 Introduction............................................................................................... 84
6.2 The Proposed Model ................................................................................. 88
6.3 Traffic Classification ................................................................................. 91
6.4 Routing Information .................................................................................. 92
6.5 Conclusion ...................................................................................................... 98
CHAPTER 7: RESULTS AND DISCUSSIONS ................................................. 100
7.1 Network Topology .................................................................................. 100
7.2 Simulations ............................................................................................. 101
7.2.1 Network Response when all nodes are active ....................................... 102
7.2.2 Network response when half of topology active ................................... 112
7.2.3 Network Response with 50 percent QoS traffic and 50 percent BE ...... 125
7.2.4 Convergence .......................................................................................... 130
7.3 Conclusion .................................................................................................... 132
CHAPTER 8: VALIDATION OF RESULTS ..................................................... 133
8.1 Introduction............................................................................................. 133
8.1.2. Validation of Data ................................................................................. 133
xi
8.1.2. Analysis of Data .................................................................................... 138
CHAPTER 9: CONCLUSIONS AND FUTURE WORK .................................... 142
9.1 Conclusions............................................................................................. 142
9.2 Future Work ............................................................................................ 143
REFERENCES ....................................................................................................... 144
APPENDIX ............................................................................................................ 152
xii
List of Figures
Figure 2.1. A Mobile Ad Hoc Network ....................................................................... 9
Figure 2.2. A Military Operation............................................................................... 12
Figure 2.3. A Corporate Network .............................................................................. 13
Figure: 2.4. Hidden Terminal diagram ...................................................................... 18
Figure: 2.5. Exposed Terminal Diagram ................................................................... 19
Figure 3.1. Key MANET Protocols ........................................................................... 29
Figure 4.1. Simulation of Three Phases ..................................................................... 49
Figure 4.2. Traffic Bytes Vs Time ............................................................................ 53
Figure 4.3. Packet ID Vs End to End Delay .............................................................. 54
Figure 4.4. Number of packets .................................................................................. 55
Figure 4.5. Traffic Bytes Vs Time ............................................................................ 57
Figure 4.6. Packet ID Vs End to End Delay .............................................................. 58
Figure 4.7. Number of packets .................................................................................. 59
Figure 4.8. Throughput Vs Mobility ......................................................................... 63
Figure 4.9. Protocol Overhead Vs Mobility ............................................................... 64
Figure 4.10. Packet Drop Ratio Vs Mobility ............................................................. 66
Figure 4.11. Throughput Vs Offered Load ................................................................ 68
Figure 4.12. Protocol Overhead Vs Offered Load ..................................................... 69
Figure 4.13. Packet Drop Ratio Vs Offered Load ...................................................... 70
Figure 6.1. Ants flow between source and destination .............................................. 87
Figure 6.2. Reactive phase of path exploration .......................................................... 94
xiii
Figure 6.3. Proactive phase of path exploration ......................................................... 95
Figure 6.4. A Sample Network.................................................................................. 98
Figure 7.1. Network Topology ................................................................................ 101
Route discovery requests ........................................................................................ 102
Figure 7.2. Hop Count while Only Class A Traffic ................................................. 102
Figure 7.3 Packet Ratio of Class A ......................................................................... 103
Figure 7.4. Hop Count while Only Class B Traffic .................................................. 104
Figure 7.5. Packet Ratio of Class B ........................................................................ 105
Figure 7.6. Hop Count while Only Class C Traffic ................................................. 106
Figure 7.7. Packet Ratio of Class C ........................................................................ 107
Figure 7.8. Topology with no QoS .......................................................................... 108
Figure 7.9. Hop Count while Only Best Effort Traffic with no QoS ....................... 108
Figure 7.10. Packet Ratio of BE with no QoS ........................................................ 109
Figure 7.11. Topology with QoS ............................................................................ 110
Figure 7.12. Hop Count, Only Best Effort Traffic with QoS active ........................ 111
Figure 7.13. Packet Ratio of BE with QoS ............................................................. 111
Figure 7.14. Upper half nodes active for class A ..................................................... 113
Figure 7.15. Hop count upper half class A ............................................................. 113
Figure 7.16. Lower half nodes active for class A .................................................... 114
Figure 7.17. Hop count Lower half class A ............................................................ 115
Figure 7.18. Upper half nodes active for class B .................................................... 116
Figure 7.19. Hop count upper half class B.............................................................. 117
Figure 7.20. Lower Half topology active for class B ............................................... 118
xiv
Figure 7.21. Hop count lower half class B............................................................... 118
Figure 7.22. Upper Half topology active for class C ................................................ 119
Figure 7.23. Hop count upper half class C............................................................... 120
Figure 7.24. Lower half topology active for class C ................................................ 121
Figure 7.25. Hop count lower half class C............................................................... 121
Figure 7.26. Upper Half topology active for Best Effort .......................................... 122
Figure 7.27. Hop count upper half BE ..................................................................... 123
Figure 7.28. Lower Half topology active for BE ..................................................... 124
Figure 7.29. Hop count lower half BE .................................................................... 124
Figure 7.30. Class A and BE nodes ......................................................................... 126
Figure 7.31. Hop count of Class A and BE.............................................................. 127
Figure 7.32. Class B and BE nodes ......................................................................... 128
Figure 7.33. Hop count of class B and BE ............................................................... 128
Figure 7.34. Class C and BE nodes ......................................................................... 129
Figure 7.35. Hop count of Class C and BE .............................................................. 130
Figure 7.36. QoS classes before Convergence and after Convergence ..................... 131
Figure 7.37. Successful Packets before and after the convergence ........................... 132
Figure 8.1. Normal Probability Graph ..................................................................... 134
Figure 8.2. Randomness graph ................................................................................ 135
Figure 8.3. Histogram ............................................................................................. 136
Figure 8.4. Diagnostics of individual classes ........................................................... 137
Figure 8.5. Summary Report ................................................................................... 138
Figure 8.6. Paired T-test between Class A and Class B ........................................... 139
xv
Figure 8.7. Paired T-test between Class B and Class C............................................ 140
Figure 8.8. Paired T-test between Class C and Class D ........................................... 141
xvi
List of Tables
Table 2.1. Protocol Stack .......................................................................................... 10
Table 3.2: Example Routing Table DSDV ................................................................ 34
Table: 3.2: Extended MANET Routing Protocols for QoS ........................................ 44
Table 4.1. Basic Factors ............................................................................................ 48
Table 4.2. Simulation Setup Table ............................................................................ 51
Table 4.3. DSDV Simulation Results ........................................................................ 52
Table 4.4. DSR Simulation Results ........................................................................... 56
Table 4.5. Parameter Table ....................................................................................... 62
Table 4.6. Load Analysis Parameter .......................................................................... 67
Table 6.1. Packet delivering strategies ...................................................................... 91
Table 6.2. Simple Probabilistic Routing .................................................................... 96
Table 6.3. Packet Formats of Forward and Backward Ant ......................................... 97
xvii
List of Abbreviations
AODV Ad Hoc On Demand Distance Vector
BE Best Effort
BRP Broadcast Resolution Protocol
CTS Clear to Send
Diffserv Differentiated services
DSDV Destination Sequenced Distance vector
DSR Dynamic Source Routing
IARP Intra Zone Routing Protocol
IERP Inter Zone Routing Protocol
Intserv Integrated Services
ITU International Telecommunication Union
LAN Local Area Network
MANET Mobile Ad Hoc Network
MPR Multipoint Relay
OTCL Object Oriented Tool Command Language
OLSR Optimized Link State Routing
QoS Quality of Service
RERR Route Error
RREP Route Reply
RREQ Route Request
RTS Request to Send
xviii
SI Swarm Intelligence
TC Topology Control
TORA Temporally-Ordered Routing Algorithm
WAN Wide Area Network
WRP Wireless Routing Protocol
ZRP Zone Routing Protocol
1
CHAPTER 1
INTRODUCTION
1.1 General Communication networks are evolving with a great pace witnessing increase in
infrastructure and applications too. Mobile Ad Hoc Network is the latest outcome in
this research. Mobile Ad Hoc Network also known as MANET [1, 2] is a network
without any available infrastructure.
There is no centralized control or any other infrastructure is needed in any MANET.
Nodes are mobile and can move whenever and wherever they want. Each node in a
MANET must be capable of functioning as a router to relay the traffic of other nodes.
These relaying nodes can themselves move too at their own even at the middle of
traffic relaying session.
Initially Mobile Ad Networks were invented for the purpose of military use but now
commercial applications of MANET have spread the use of MANET in many other
areas too.
1.2 Problem Statement Specific and dynamic nature of MANET requires special routing functionality than
methodology used in routing protocols available for internet or wired network. There
have been proposed a large number of routing protocols for MANET by research
2
community. Some of these are derived from old traditional approaches tailored for
MANET while some are specially invented for MANET. Rich applications like
multimedia require more routing functionality than only basic available services.
Therefore provisioning of quality of service in such MANET is a hot issue due to
specific nature of such dynamic networks.
Research is being carried out in Quality of Service provisioning in MANET but the
work is still considered in its infancy stage. Two main streams are reported in research
literature for providing said quality of service in mobile ad hoc networks; one deals in
proposing extensions to already proposed best effort routing algorithms while the
other is dedicated or purpose built QoS aware routing protocols. A thorough study and
analysis of these conventional approaches has revealed ineffectiveness of these
techniques.
Dynamic nature of MANET does not guarantee of route availability during any
session of particular application. Real time and multimedia applications demand
availability of route during whole dialogue session. One of the techniques for this
could be through already searched alternate routes.
Conventional routing approaches are found heavily dependent on feedbacks from
other layers. Many of the stale routes in MANET remain exist in caches even while
not being used; therefore avoiding these routing entries is another requirement better
to be met in MANET.
3
Swarm intelligence (SI) based optimization techniques [3] have shown promising
solutions to communication networks. Some protocols are being developed for routing
in Mobile Ad Hoc Networks too.
Therefore there is a need to propose a novel QoS aware routing protocol for Mobile
Ad Hoc Network based on swarm intelligence technique.
1.3 Objectives and Scope Getting motivation from Swarm Intelligence based routing, Ants inspired traffic
classification is proposed for Quality of Service provisioning in MANET based on
stigmergy of ants. It is a swarm intelligence (SI) technique involving natural ants like
optimization solution, where stigmergy is the process of indirect communication
between agents through modifications induced in environment.
The Objectives of this research work are as follows:
To study routing protocols proposed so far in MANET through literature
survey.
To prove ineffectiveness of conventional approaches through performance
evaluation of leading routing protocols for MANET through simulation.
To investigate Quality of Service provisioning in MANET
To analyze the phenomenon of Swarm Intelligence and its applicability in
Mobile Ad Hoc Network environment.
4
To propose a novel solution for providing Quality of Service in MANET
through swarm intelligence based technique using Ants optimization
methodology.
To provide an adaptive solution in case of degraded quality of service by
instantly making the alternate routing paths available through meeting
acceptable Quality of Service criteria.
To develop a layer independent model, which should not rely on any feedback
information from other layers as opposed to other conventional algorithms.
To eliminate stale route entries in routing tables through evaporation property
of pheromone in Stigmergy of Ants.
1.4 Original Contributions Work in this thesis has contribution in the field of Mobile Ad Hoc Network where
provisioning of quality of service is still an issue. There are numerous routing
protocols proposed in the literature much are yet to reach a mature level for every
scenario available for best effort services which can not yield good results for rich
applications like real time and multimedia demanding other important metrics. Quality
of service provisioning is done through swarm intelligence found in ants. Traffic is
classified in four classes. Main contributions of our research are as follow:
Prioritizing traffic is not a simple issue in MANET rather a complex one due to
dynamic nature. Therefore Swarm intelligence based ants inspired traffic
5
classification is achieved in this research work showing the promise of the
technique in MANET.
Usually two routing approaches are used in MANET namely Reactive and
Proactive approaches. This research work is hybrid of both the techniques, in
which reactive approach is used in initial route exploration while proactive
approach is used to make alternate routes available in case of path loss, which
is a continuous process during data session.
Stale routes deletion is widely recognized as a problem in MANET.
Evaporation property in MANET is exploited well here to avoid stale route
entries in routing tables.
1.5 Organization of Thesis This thesis is organized into a total of nine chapters, reflecting research experience
built up during the course of work carried out in provisioning of quality of service in
Mobile Ad Hoc Networks. Chapter one gives a brief introduction of providing quality
of service in MANET along with problem statement and objectives of this research.
Chapter two and three discuss in detail the literature survey about Mobile Ad Hoc
Networks, its applications and special characteristics, routing in MANETS and
Quality of Service provisioning. Chapter four presents the performance evaluation of
two leading MANET routing protocols DSR and DSDV coupled with simulation
results. Chapter five gives introduction of Swarm Intelligence concept and its
applicability in Mobile Ad Hoc environment.
6
Chapter six presents the proposed solution of Ants Inspired Traffic Classification in
Mobile Ad Hoc Networks. Chapter seven includes the discussion on detailed results
achieved. Chapter eight presents the validation of results on the basis of ANOVA
tests. The final chapter nine concludes the thesis and gives directions about future
works in this connection.
7
CHAPTER 2
TECHNICAL BACKGROUND In this Chapter, a thorough literature survey has been carried out to investigate Mobile
Ad Hoc Networks characteristics, its applications, routing and quality of service
approaches proposed in the literature.
2.1 Mobile Ad Hoc Networks In this era of technological advancement the improvements in wireless communication
and the latest development in portable computing devices have enabled to
communicate with ease anytime and anywhere. One can observe users moving around,
while simultaneously keeping connectivity intact with all the rest while traveling. This
is termed as mobile computing which has established a great deal of interest in these
days.
Generally, there is required only one hop connection to the fixed network by most of
the contemporary computing applications. This is the example of famous cellular
network model which requires the installation of huge base stations with antennas and
access points for wireless communication to take place. Therefore like wired network
such networks rely on infrastructure to be installed for communication between
network nodes.
The availability of wired network or infrastructure based wireless can not be assured
the communication possible everywhere every time, for example natural disasters and
8
such other emergencies demand a rapid deployment without any available
infrastructure. Cost and terrain conditions again limit the use of above wired or
wireless but based on infrastructure. A Mobile Ad Hoc network is suggested for these
scenarios. As the name suggests that it is Mobile and Ad Hoc, means movement is a
special characteristic along with adhocism.
A Mobile Ad Hoc Network (usually called MANET) [1, 2] is the network of laptops,
palm devices, mobile phones, or any other possible devices capable of communicating
with such other devices in the network. No centralized control or any other
infrastructure is needed in any MANET, this means that this is unlike other wireless
networks consisting on access points or base stations or tall antennas. MANET devices
rely on their own small attached antennas. Mobile nodes in MANET can move
arbitrarily. Each node in a MANET must be capable of functioning as a router as there
are not available any dedicated routers, gateways or switches. Communication
between mobile nodes in MANET takes place through participating nodes, which
relay the traffic of other nodes passing through them. These relaying nodes can
themselves also move any time at any speed they like [4, 5]. A typical MANET
picture is shown in Figure 2.1.
9
Figure 2.1. A Mobile Ad Hoc Network
2.2 Mobile Ad Hoc Networks Protocol Layer Stack Discussion of any network model remains incomplete if model is not discussed in the
way of OSI layers. Table 1-1 shows the protocol stack which consists of the five
layers which are physical, data link, network, transport and application. It has
similarities to the TCP model, as OSI first three layers are merged into one layer, the
application layer.
It is the Network layer which distinguishes the MANET from the TCP/IP model. Here
the ad hoc network routing protocols are proposed. In the second layer which is data
link layer any common wireless MAC layer protocol can be used.
10
In the transport layer, the TCP or UDP can be used, but it is found that TCP is not
efficient in performing well in such networks, rational behind this is that, in wired
networks the dropped packets are mostly because of congestion while in mobile ad
hoc networks environment this is due to some other factors like as transmission error,
route change, etc. Application layer works same as in TCP/IP that is depending upon
the requirement in application concerned.
The work in this thesis relates to only the network layer and since this is a layer
independent approach, therefore in the subsequent sections, the discussion would only
be focused to the network layer instead of all layers.
Table 2.1. Protocol Stack
OSI Layers TCP/IP Layers MANET Layers
11
Although mobile ad hoc networks (MANETs) were initially invented and used for
military purposes but rising popularity and commercial usage of MANETs, has
revealed much wider applications of such networks.
2.3 Mobile Ad Hoc Networks Applications Numerous areas are there, where applications of Mobile Ad hoc network can be found.
Some of them are discuss below [6].
2.3.1 Applications in Natural Disasters
In relief operations there can not be any other choice than a MANET. Natural
disasters are harder to predict, therefore no prior planning or any infrastructure can be
made available in these situations. Outcome of such catastrophes result usually in
complete havoc to any installation. Therefore setting proper infrastructure is even not
possible, but very expensive and very difficult to deploy. Therefore the only choice
left is of an Ad Hoc Setup. Following can be achieved by deploying an Ad Hoc
network.
Writ of Law
Efficient relief operations
Rescue operations
Reporting to back office
Efficient search and locate operations
12
2.3.2 Battlefield Scenarios
Some of the essentials requirements of a battlefield assignment include network ease
of deployment, secure communication, IP based protocol, Continuous connectivity in
high mobility and anti jamming techniques. In much of the scenarios, combat acts are
often spontaneous i.e. with no prior predetermined network infrastructure available. At
other hand, the military should be capable to form a network at its own where and
when this is required as shown in Figure 2.2. Initial applications of MAENTS were of
course in battlefields.
Figure 2.2. A Military Operation
13
Combat operation strategies are designed for irregular terrains therefore making
geography an important parameter for consideration. It is not feasible for military
personnel to acquire any GPS information openly therefore coordinating with each
other for such information gathering is highly beneficial.
2.3.3 Enterprise Networks
The Enterprise networks are very useful in connecting various premises of same
organization. Employees often move freely and arbitrarily while still enjoying the
connectivity. Diagram 2.3 depicts these possible scenarios in a detailed manner.
Figure 2.3. A Corporate Network
14
A good example is of a conference, meeting or a seminar, where the speaker can get
advantage of multicasting the presentation slides to the intending participants. A
notice board shared among participants can serve in question and answer sessions
during the meeting.
Real time applications are much possible in wired network or current internet.
Therefore support for such applications in MANET is being researched but still in its
infancy stage. This research involves the provisioning of this quality of service in
Mobile Ad Hoc network.
Quality of service is considered a challenging task rather conventional routing in
MANET is not that simple as in wired network or in internet. This is because of the
special properties of Mobile Ad Hoc Networks. Following section elaborates briefly
such characteristics.
2.4 MANET Characteristics In this section a brief description is presented about the major characteristics of highly
dynamic mobile ad hoc network. These characteristics are important to understand the
main requirements of designing the protocols for such challenging environment of
Mobile Ad Hoc Network.
2.4.1 Scarce radio bandwidth
Mobile ad hoc networks operate on wireless links which use scarce radio bandwidth.
In view of the fact that radio channel represents a prelude for communication between
15
mobile nodes in wireless environment, inadequate bandwidth represents a basic
setback and so, insists on the demands on a much well-organized approach for the
distributed use of resources and management of the network. Misuse of this essential
network may lead to the multiplication of packets collisions and further to the
incapacity to send and receive packets.
2.4.2 Distributive and without any infrastructure
In MANET all operations are distributive in nature, this in contrast to any centralized
controlled approach. Nodes are dependent on each other to exchange packets and
make communication possible. Here is present, no central network infrastructure to
configure any other network operation. This means that it is unlike conventional
cellular network which requires some infrastructure like base stations and tall towers
etc.
2.4.3 Dynamic topology and mobility Topology is a major issue in Mobile Ad Hoc Network as nodes move arbitrarily at
their own causing frequent changes in network topology. So there exists no fixed
topology in any MANET. This dynamic topology poses serious challenges to any
designing strategy.
2.4.4 Little contact duration among nodes
Sustaining connection between nodes is an important issue in MANET. Due to
mobility and other factors no communication session can be guaranteed as node
16
contacts may break often. A good example in this regard can be of ad network of
traffic where traffic is moving.
2.4.5 Economical solution
Mobile Ad Hoc Network provides economical solution in terms of cost of deployment,
because here cost of cabling and other infrastructure is not required. Communication is
carried only among nodes involving only each other at their own without involvement
of any base station or any other communication supporting equipment.
2.4.6 Beyond LOS connectivity Line of sight connectivity is an issue in any wireless communication setup. A mobile
Ad Hoc Network provides beyond line of sight connectivity through its multi hop
routing technique so that by coordinated efforts of far away nodes, communication is
made possible through multi hop routing.
2.4.7 Numerous nodes and huge network
So many nodes sparsely spread over large terrain pose a serious challenge in any
designing strategy. This problem coupled with high mobility rate turns into even a
bigger challenge.
2.4.8 Bit error rate
Bit error rate is an inherent problem of Mobile Ad Hoc Networks due to links
operating on wireless channels. Signal reception at receivers is often with bit errors
17
due to wireless own characteristics. So there must be design strategies to deal with this
menace.
Finally it is concluded that in short, Mobile Ad hoc networks are:
o Self-creating
o Self- configuring
o Self- administrating
o Self- organizing
2.5 MANET Challenges After discussing the potential applications and benefits of mobile ad hoc networks it is
pertinent to mention some challenging areas too [7]. These are discussed in detail in
the following subsections:
2.5.1 Dynamic Topology
Mobility in MANET gives rise to its dynamic nature, where dynamic and ever
changing topology is a major challenge for the one who designs the network. It
requires manipulative and improvement of network layer and data link layer strategies
to deal with the high mobility in network.
2.5.2 Asymmetrical Links Asymmetrical links problem is special to wireless networks; in Mobile Ad Hoc
Networks also, links are usually not all at the same characteristics and therefore
routing strategies must take care of this asymmetry.
18
2.5.3 Frequent link breakage
Due to high mobility, frequent disconnections and partitions occur in the network
accordingly. Therefore this again is a major issue that should be dealt appropriately.
2.5.4 Hidden terminal and exposed terminal problem Exposed terminal and hidden terminal problem are inherent wireless carrier sensing
problems. These problems can well be understood by following two example
scenarios.
(a) Hidden Terminal
In following scenario in Figure 2.4, node B can communicate with both the nodes A
and C, while node A is out of sight of C node. Therefore any transmission from A to B
is hidden from C. Hence during such transmission if C starts transmitting to B then
collision will occur at B.
Figure: 2.4. Hidden Terminal diagram
19
(b) Exposed Terminal
The exposed terminal problem happens when a transmission is being carried out and
the next transmission is deferred considering the channel as busy though the second
transmission is possible without interference. In the following scenario in Figure 2.5,
nodes C can talk to node B and node D both, while node B can talk to node A and
node C both.
Now for example if node B is transmitting to node A and if node C have something to
transfer to node D then this transmission is deferred believing the channel as busy and
interference may happen. Now looking at Figure 2.6 it is very much clear that node D
can have the successful reception from node C without interference as it is out of
range of A and B dialogue session so the transmission is very much possible but is not
executed due to exposed terminal problem.
Figure: 2.5. Exposed Terminal Diagram
20
2.5.5 Security issues
Design considerations are also needed in security matters in Mobile Ad Hoc Networks
as these operate on wireless links which may suffer security issues very much related
to these wireless links.
2.5.6 Distributed operations Distributed operation is of utmost importance in MANET as there is no central
controlling and there should also be finest usage of precious resources as bandwidth,
power and etc.
2.6 Quality of Service Here quality of service is explained as the combined cause of overhaul performance
that establishes the extent of fulfillment of the requirement of a user of that service.
The network is anticipated to promise a set of measureable pre defined service levels
to the user of that service, the commonly used metrics for measuring this quality of
service are:
(i) Bandwidth
Bandwidth is very important in providing quality of service with respect to the
application requirements. Rich applications like Streaming video require more
bandwidth as compared with the normal HTML browsing. Bandwidth allocation
should be in such a way that rich application may be entertained while at the same
time starvation of other application is also avoided.
21
(ii) Packet loss
Packet loss is another important parameter in providing quality of service in the
network. Some applications are sensitive in delivery of packets such that if some
packets are lost, then the integrity of the application is compromised which is not
acceptable. Packet loss occurs in the network, especially in the case of wireless
channels.
(iii) End to end delay
Some applications are there which are sensitive to the delay associated in the
transmission. These applications should be served with acceptable delay limits in
order to achieve quality of service for such type of applications. For example, real
time applications want to be reached at the destination within reasonable time limit.
(iv) Jitter
Jitter is basically the variation in delays. Some packets experience delay of some value
while other packets of same application are subjected to another delay bounds. This is
especially cumbersome in case of real time applications.
In this research, the proposed solution caters for the first metric which is the
bandwidth in a way that traffic is prioritized according to the bandwidth requirements
of the respective applications so that more bandwidth hungry applications are
entertained by providing more bandwidth.
In simple terms, QoS provisioning can be visualized by widening the communication
pipe [8]. In wire based network the quality of service can be provided in two ways [9]:
22
Over-provision of resources
Network traffic Engineering
Through over-provision of resources, we add plenty full resources but as evident all
the users remain at one service class. In network traffic engineering, users are divided
in classes according to their application requirements and each class is given a defined
priority.
Network traffic engineering has two main types [10]
(a) Reservation based engineering
This is also called Integrated services (Intserv) where resources in the network are
beforehand reserved for particular demanding application, that is why this is a method
of reservation based engineering. This is dependent on another important protocol
called RSVP which is an out of band signaling protocol.
(b) Reservation less Engineering
This is also called Differentiated service (Diffserv) where no resource reservation is
carried out in the network, rather per hop behavior is done to each packets need. This
is definitely lightweight on inner nodes as no longer resource reservation sessions are
maintained here.
23
2.6.1 Challenges for providing QoS:
Due to the specific nature of MANET, QOS provisioning has been widely recognized
as a challenging task [11, 12]. Therefore for providing quality of service in MANET,
following issues need to be handled. These issues are currently being addressed
through research but the work at present is still in its infancy stage [13, 14].
(i) Wireless medium
Unpredictable nature of wireless medium does not guarantee the link availability and
stability. Links are not always available as bidirectional, rather any link may become
unidirectional any time.
(ii) Decentralization
Mobile Ad hoc networks are not centrally operated in nature, so handing over a central
regulator in maintaining the network operations and any other possible reservation is
not suitable.
(iii) Broadcast nature
Because of the broadcast characteristic of radio transmission, abundant hidden as well
as exposed terminals may show up, which turns reliable end-to-end delay and
transmission control very complex.
(iv) Mobility
Due to mobile nature of nodes in MANETs, the topology is highly dynamic and life
time of a link is very short and unpredictable.
24
(v) Power consumption
Nodes in MANETs usually operate on limited battery powers and hence power
consumption becomes another serious issue in mobile ad hoc networks.
(vi) Complex routing
The routing problem becomes more complex because of the wireless nature and
mobility in MANETs. The random and rapid movement of the nodes and the uncertain
wireless links results in incapability of traditional routing protocols.
Further, there are partial and over reservations which are specifically related to QoS in
MANETs that contribute to the above mentioned challenges.
The main internet quality of service approaches Intserv and Diffserv cannot be directly
applied in MANETs [15] because of the following reasons.
Integrated services are difficult to apply in resource limited MANET. Tiny and less
capable nodes in MANET cannot afford huge process requirements of integrated
services, as resource reservation protocol like RSVP requires heavy processing on
intermediate nodes. As bandwidth is limited in the Mobile Ad Hoc Networks,
therefore its usage should be minimized. This is contrary to the case of RSVP protocol
which requires out of band signaling. RSVP also requires scalability support when
number of flows increases, which again is an issue in MANET.
Providing quality of service on differentiated services basis in MANET is again not a
simple and easy task. Though differentiated services model is an easy one but still
25
there are issues for its implementation here. For example marking a node as core,
ingress or egress in dynamic topology requires complexity.
Implementation of service level agreements is again a complex task in an entirely ad
hoc topology where none of the nodes takes responsibility of providing some
assurance of any sort.
2.7 Conclusion In this chapter an introduction about MANET has been presented. Applications of
Mobile Ad Hoc Networks have been discussed in detail including special
characteristics of such networks and challenges in MANET. An introduction about
quality of service is presented coupled with its provisioning in such dynamic ad hoc
environments. This chapter concludes that quality of service provisioning can not be
achieved by the methods applied in the prevailing wired networks.
In the next chapter, routing approaches in MANET are presented. Quality of Service
aware routing through current techniques is also discussed and drawbacks associated
with these are also mentioned.
26
CHAPTER 3
ROUTING RELATED WORK
3.1 General Keeping in view resource limitation and highly dynamic topology, extensive research
has been conducted in the routing area of MANETs and a plethora of routing protocols
exists for MANETs.
These routing protocols wrap a broad variety of propose choices and techniques, from
plain amendment of fixed network protocols, to further intricate several hierarchical
strategies. Several of these routing protocols are developed depending on comparable
sets of best guesses [16].
On the basis of routing strategy used, these protocols may broadly be divided into
three main classes namely, Proactive routing strategy, Reactive routing strategy and
Hybrid routing strategy.
Proactive strategy for ad hoc networks is resultant from the conventional distance
vector and link state protocols designed for usage in the fixed internet. Basis of these
table driven protocols can safely be claimed as well known Bellman-Ford Algorithm
[17]. The basic theme here in proactive strategy is that each node keeps record of path
to all other nodes in the network every time, which is achieved through routing tables.
So routing table updation is another important event here which is handled either by
updates in two forms event-triggered and periodic.
27
The periodic updates are done by exchanging routing information only at defined time
intervals among nodes. These updates occur at specific intervals and do not take into
account the traffic characteristics and mobility in that particular network. Possible
events for triggering can be sudden route breakage or addition or removal of any node
in that network.
Proactive strategy has the benefit that routes are immediately available at the time of
need. Since every node constantly keeps a current route to all of other nodes present in
the network, therefore a node just checks its routing table when there is a need of data
transmission to a destination and begins the session of data communication instantly.
It is worth noting here that the main weakness of these routing strategies is that the
control overhead is considerable in huge networks and in networks where mobility is
random and fast [18].
In short most of the Proactive routing protocols are the same old traditional wired
LAN/WAN protocols tailored towards mobile ad hoc networks [19], hence these
protocols consume larger bandwidth for frequent updates in a highly dynamic
topological network like MANET. Since the bandwidth is at premium in wireless
network and hence its usage must be minimized therefore a Proactive routing is not
the best candidate for a routing algorithm in MANETs. Some of the proposed
proactive protocols are DSDV, WRP, OLSR and etc.
28
On the other side, reactive routing protocols adapt a very different strategy of routing
[20]. As in proactive protocols, each node always needs to keep a routing path to each
other node in the network, therefore huge overhead is observed here. In a fixed
network, where scenarios of connecting links are rarely altered and plentiful resources
are there, managing the complete connectivity graphs is not a worthless attempt. The
main advantage is that at the time of need, a routing path is available at the spot [21].
In contrast to this link patterns change more often in a mobile ad hoc network
therefore control overhead is more over there.
In Reactive Protocols, when there is need of data packets transmission to some
destination, routing table of the source is consulted to find whether there exists a route
or not. If no routing path is discovered then route discovery procedure is started. This
is usually done by flooding the Route Request Packets in whole the network. In order
to minimize this overhead, a number of optimization techniques are applied to the area
of search in the network. The main advantage in this routing technique is that
overhead involved in signaling is probable to be minimized as compared to
approaches involving proactive strategies, mainly in networks where the load of traffic
is small to reasonably medium. In the scenarios where the communication dialogues
are more than medium, then this control overhead initiated by the route discovery
requests becomes high, and also may even exceed beyond the routing overhead
involved in proactive strategies. Some of proposed reactive routing protocols are
AODV, DSR, TORA and etc.
29
Though Reactive Routing Protocols minimizes the bandwidth usage but offers delay
as when a packet needs a transmission a global search is conducted before actual
transmission, this is against the already available routing tables in proactive protocols.
Therefore in some manner, reactive protocol plays good while in other situations we
need the proactive protocols [22].
The third type on the basis of flat routing schemes of proposed routing protocols is the
blend of both the proactive and reactive routing protocols and hence is called hybrid
routing protocol. These hybrid protocols may exhibit proactively in some set of
conditions, at the same time demonstrate reactive behavior in some other set of
conditions. Such routing strategies tolerate for elasticity based on the distinctiveness
of the network. Some of the hybrid routing strategies are the Zone Routing Protocol
and also Distance Routing Effect Algorithm for Mobility. Following Figure 3.1 shows
some of the MANET Routing Protocols.
Figure 3.1. Key MANET Protocols
Ad Hoc Routing Protocol
Table Driven Hybrid On Demand
DSDV OLSR
WRP
ZRP AODV DSR TORA
ABR
30
A thorough study of these protocols is very important while doing research in
MANET related issues. A detailed survey of some of these protocols is presented in
section 3.2.
3.2 MANET Routing Protocols There is variety of proposed routing protocols for MANET which are reported in the
literature. In this section, we discuss some of the more popular and leading protocols
which have caught more attention.
3.2.1 AODV (Ad Hoc On Demand Distance Vector)
This is the leading routing protocol proposed so for in the category of on demand or
reactive routing protocols. Unlike table driven protocols, it does not maintain status of
the network via continuous updates [25]. This approach assists in minimizing the
flooded messages and also size of routes tables. It was designed after a distance vector
routing protocol DSDV but is much efficient than DSDV. Actually AODV is a blend
of DSDV and DSR. It has the actual on-demand technique of discovering the route
and also route maintenance from DSR, but uses sequence numbering and also the
periodic beacons of DSDV. New routes are found through the process of RREQ and
RREP where RREQ packets are broadcast and RREP are unicast in nature. While
route maintenance uses RERR packets for remedy of route breaks, routing information
is kept afresh by the usage of sequence numbers, which is the idea borrowed from
DSDV [26].
31
Whenever there is a packet received from the upper layer for onward transmission,
then the temporary cache is checked for if a routing path is available for this intended
destination. If no entry is found then Route Request is generated and these RREQ
packets are broadcasted in whole of the network. These RREQ packets are listened by
every node in the network and the destination or any other node which has information
about route for the said destination may respond with route reply packet generally
known as RREP packet. This RREP is embedded with the sequence number which is
equal or larger than one contained in RREQ packet. Sequence numbers are very
important in Ad Hoc On Demand Distance Vector to avoid loops and redundancy. In
this Protocol, the already entertained requests are simply discarded, data transmission
begins after successful processing of Route Reply packet. Afterwards good routes can
be selected by receiving higher sequence number Route Replies. Time out is another
important feature in AODV protocol as after the transmission means if routes are no
more needed, they are deleted. RERR is the Route error packet produced whilst a
break in route is detected. On reception of RERR packet the transmitting node may
take appropriate strategy of route discovery if it has not yet finished previous data
transmission session.
In cache of the every relaying node following information is kept during data session.
Sequence number of destination
IP address of destination
Hop count: hops to the destination given by destination IP address.
32
Next hop: next node IP address to go in order to reach the destination given by
the destination IP address.
Precursors: IP addresses of every node that use this node as next hop to their
destination.
Lifetime: deletion time of the route
Hello messages are special to AODV only. AODV uses this message to check the
status of neighboring nodes. Whenever a node fails to transmit ‘hello’ packet within
the prescribed time limit it will be marked as not active node and the link is not
working. The neighbors under these circumstances are responsible to inform every
relevant node about the not active node.
Every route table entry must contain the most recent information regarding destination
sequence number for that route to be a valid one. Any particular node updates in turn
this sequence number in one of two conditions.
Whenever a node broadcasts request, it increments its sequence number.
When a node unicast reply back towards the source node. It increases its
sequence number.
Likewise, nodes also update destination sequence entry, either on receiving a sequence
greater then recorded one in its routing table or on receiving route error message
indicating that routing path to that particular destination has been not working or no
more available.
33
When a node receives a data packet that node determines whether this is the node as
destination. If it turns to be destination, the packet is effectively received at this
destination node. On the other hand, it starts looking up the next possible node to the
intended destination in the routing table and it unicast the data packet for further relay.
If deliverance turns unsuccessful, the RERR packet is generated and sent back.
Receiving this RERR packet the destination node removes the relevant path and
initiates the new discovery procedure afresh.
3.2.2 DSDV (Destination Sequenced Distance vector)
The DSDV [27] is a Proactive routing algorithm based upon well known classical
distance vector algorithm of Bellman-Ford. Routing tables are maintained and updated
accordingly, so broadcast periodic routing table update packets consume the
bandwidth. So the main weakness of DSDV is that when network grows these packets
also increase. The main improvement here to Bellman-Ford algorithm is loop freedom,
which is made possible by assigning sequence number to each entry in routing table,
which avoids stale routes.
Every network node keeps up a table, which contains the address of the destination,
the lowest number of hops to reach that specific destination, the next hop along the
path of that particular destination and the sequence number. A model routing table in
DSDV at a Mobile Host is presented in Table 3.2.
34
Table 3.2: Example Routing Table DSDV
Destination No. of Hops Next Hop Sequence no.
Host 1 2 Host 2 S1
Host 2 1 Host 2 S2
Host 3 2 Host 2 S3
In highly dynamic environment of ad hoc network DSDV suffers from bandwidth
scarcity due to larger routing tables which in turn also tends to slow network
convergence too.
Routing table entries are advertised in Destination Sequence Distance Vector Routing
Protocol. These entries usually contain, destination address, number of hopes to this
destination and importantly the sequence number. This sequence number is matched to
have the latest one discarding all previous. On further relaying this entry the
corresponding node increments the hop count to add its own.
Sequence number generation is subjected to a new broadcast or on detection of a
broken link. Fresh Route is chosen which has latest sequence number and older
sequence number routes are deleted. If there is a tie of sequence numbers, then the
route chosen is that which has better metric.
Concluding the discussion on DSDV it can safely be said that, the main advantage
along with proactive routing in DSDV is the Loop freedom. There are several aspects
where efficiency of DSDV is low, these include:
35
Finest parameters in DSDV akin to settling time and etc. are not simple to
resolve. The aftermath of this is redundant utilization of bandwidth.
Overhead is much more due to plenty of regular updates.
DSDV is not able to sustain multi-cast routing.
Scalability is again an issue here as when the network grows, the routing table
also expands. Every node in the network is needed to keep routing path
information of all other nodes present in the network, although this information
is not needed for current data sessions.
3.2.3 DSR (Dynamic Source Routing)
The Dynamic Source Routing (DSR) [28] is an on-demand or reactive routing
protocol. Therefore unlike other proactive routing protocols, DSR involves no updates
of whichever type at any stage inside the network. The DSR uses Source routing for
forwarding data packets which distinguishes DSR from other reactive routing
protocols. It is lightweight on inner routers as due to source routing, the routing
information keeping is not needed at every host. The sender becomes aware of
complete destination address before transmission and appends this address in header
of the routing data packet at the beginning. It is loop free due to source routing.
Extensive use of cache and promiscuously listening are the main optimizations to DSR
when network is at low mobility.
Route finding and Route maintenance are the two major mechanisms of DSR [29].
Route discovery mechanism is initiated when there arises, a need to transmit packets
36
to unknown destination. Route maintenance mechanism is applied to observe the
health of discovered routes and then to kick off route discovery on failure of a routing
path. DSR does not need any Neighbor discovery as no need of intermediate nodes to
keep record of other nodes because source node itself provides the whole hop by hop
routing path to the intending destination of current session.
Whenever a node in the ad hoc network wishes to transmit a packet of data to an
intended destination and it does not have a route to that destination, it initiates a route
discovery process to vigorously establish this routing path. This is again flooding of
Route Request (RREQ) packets as in AODV. Node receiving this RREQ packet,
performs two tasks, first is to check itself as destination or if it has route to destination,
second is performed in failure of first task so the RREQ packet is forwarded.
Ultimately the RREQ packet reaches at the intended destination or node having the
destination address, so it replies as RREP packet which again is source routed as now
the node knows complete reverse path. Now at the end the initial sending node
receives the whole path hop by hop to this destination which it stores in its cache.
Medium Access Control Layer and Network Layer are responsible for detecting any
link breakage, depending upon the version of DSR Protocol used. This is done
throughout by the extensive use of acknowledgements received commencing at the
corresponding layer.
If at some point of time any path on a source route is busted during the course of data
communication session, a Route error (RERR) packet is sent to the first initiating
37
node. The first task which this initiating node does on reception of any RERR packet
is that it deletes this information from it cache to prevent further data packet loss. Now
if data session is pending, then a new route discovery process is started as previous
initial one. Intermediate nodes also delete the broken link entry from their cache to
avoid any further use of the said routing path.
Nodes in DSR promiscuously hear the data sessions of other nodes to judge if it has a
better route for this session. If it finds this, then it sends the source node about this
better route. So this is an added advantage in DSR of multi paths through promiscuous
learning. Therefore using source routing the DSR protocol can access significant
amount of routing paths as many intermediate routes are also known by default.
By making use of route cache, DSR becomes eligible to respond all requests of routes.
Therefore many other routes to the destination are also found which may help in any
route breakage.
Enjoying access to several alternating routes minimizes the flooding of route
discovery, which often is a bottleneck in these networks. Many of the proposed
specifications of DSR does not include any clear technique to retire stale routes
present in the cache, or to prefer the fresh routes when coming across a choice with
multiple paths. Although a few of these stale routing entries are deleted through
packets of route error, but due to node mobility and listening promiscuously, it may
happen that these caches are again filled with such stale routing entries rather than
having deleted by route error packets.
38
3.2.4 OLSR (Optimized Link State Routing)
The Optimized Link State Routing (OLSR) Protocol [30] is a Proactive Protocol. This
is based upon Link State Protocol (Old Dijikstra Algorithm). The OLSR is basically a
table driven routing protocol therefore routing tables are maintained using link states
as metrics. The main idea is to reduce flooding by using Multipoint relays (MPRs).
The MPRs are nodes which are selected used to minimize the broadcast packet
transmissions and also the volume of periodic update packets in the network. Hello
and Topology messages are special to OLSR used for discovering the link state
information. Shortest path is selected using this topology information to calculate next
hop to the desired destination.
OLSR works on the basis of Open Shortest Path first but is tailored to Mobile Ad Hoc
Network environment. Main aim of such tailoring is to optimize the flooding process.
Special hello messages are used here to discover first hop neighbors. Multipoint relays
(MPRs) are chosen on a set pattern, so that routes to all second hop neighbors are
available through these MPR nodes. Storing and forwarding of TC messages is done
through these MPR nodes which, contains information about who selected these as
MPRs. The information about forward path for TC messages is not common through
all the nodes but fluctuates due to the source, merely a division of nodes store link
state detail, here only those nodes are advertised which denote MPR selectors. This is
in contrast to the broadcasting of all link information.
39
Path calculation is done at each individual node using shortest hop route. Therefore
routing tables are designed having all the paths to every one of the destinations in the
network.
Finally it can be concluded that proactive routing strategy used in OLSR has routing
information available before hand, therefore no delay as is in the case of on demand
protocols. Flooding is minimized as not in many proactive protocols. OLSR is
predominantly suited for thick networks, and when the network is thin, it means less
number of nodes than each neighbor of every node that acts as a multipoint relay; the
OLSR then behaves just like clean Link State routing protocol.
3.2.5 ZRP (Zone Routing Protocol)
The Zone routing protocol [31] is a hybrid algorithm, which blends both the properties
of proactive and reactive routing protocols while getting advantage of both the
techniques. The beauty or merger of both proactive and reactive is in such a way that
every node maintains its zone which consists of all nodes inside the defined zone
radius. Remember that this zone radius is measured in hops and can be varied
accordingly [32]. Inside this zone, routing tables are maintained just like proactive
strategy while outside this zone routing information is achieved same like reactive or
on demand strategy. Here Hass et all described the inside proactive as Intrazone
routing protocol (IARP) and Interzone routing protocol (IERP) for reactive part of the
ZRP. IARP can be link state routing or distance vector routing which depends on
execution of the algorithm.
40
Routing tables are maintained inside the zone so routing information readily available
here. When a node intends to send data to a destination outside the zone it initiate the
same on demand type route discovery process which is Route request (RREQ) packets
are flooded but outside the relevant zone. This flooding is broadcast by the nodes at
the boarder of the said zone called Broadcast Resolution protocol (BRP). So flooding
is minimized by boarder nodes only.
Mobile Ad Hoc networks are different than other networks in link breakage. As link
goes down here due to high mobility, so route maintenance is very essential in
MANET. Route maintenance is easy in intra zone, and in outside zone the source
routing strategy can be adopted accordingly.
Finally about ZRP, it can be said that the Zone Routing Protocol is better in using less
number of Route request messages but the algorithm is difficult to in implementation.
Main drawback appears when the network becomes large and as a result more
overhead is generated. Address resolution is another issue of concern in Zone Routing
Protocol.
3.3 QoS Aware Routing None of the above mentioned protocols take into account the Quality of Service rather
these are only best effort delivery protocols in nature. Where QoS is a set of service
requirements that needs to be assured by the network during the course of transferring
a packet flow from a source to its destination therefore, the primary goal of the QoS-
41
aware routing protocols is to determine a route from a source to the destination which
must guarantee the requirements of such demanded QoS level. The nature of MANET
does not guarantee such an assurance. Hence providing quality of service aware
routing in MANET is a more complex issue.
Research is being carried out in developing QoS aware routing protocols and some
protocols have been designed in this connection but yet to reach an efficient and
mature solution in this regard. The most common metrics to attain quality of service in
MANETs are bandwidth, delay, jitter, packet loss and etc.
Some research is being carried out by proposing some extensions to the already
available MANET best effort routing protocols to achieve certain amount of quality of
service.
Reactive routing protocols are extended in such a way that the Route Request packets
also carry along with themselves the desired metrics. While proactive protocols are
modified in a way, that their routing tables also contain the link status in terms of
quality of service metrics.
The basic AODV protocol is extended to provide QoS support in Mobile ad hoc
networks [33] in the following manner. During route discovery sessions Route
Requests and Route Reply packets are appended with extension messages which
specify the required metrics.
42
A node is only tuned as a valid hop of the route if and only if it can meet the specified
requirements in the RREQ packet. If the Route has been already established and the
QoS specified requirements cannot be met any longer, then the node originates an
ICMP QOS_LOST message and sends it to the initiating source.
Here two strategies are adapted for assuring minimum bandwidth and maximum
possible delay. In the case of delay, in response of a Route request packet that node
subtracts the value in Route Request that time required by that node in processing. The
packet is discarded when this turns negative.
In case of Bandwidth, a comparison is made between the demand of Route Request
and available link capacity. The packet is discarded if RREQ demand turns higher
otherwise RREP is generated accordingly.
I. Jawhar [34] presented some extensions to some of the mentioned best effort
MANET routing protocols. It was claimed that DSR is many times the basis for many
of these extended QoS aware protocols. Extension for QoS metric is added by source
in Route Request packets to further exploit the source routing capability of DSR as
against to other MANET routing protocols.
YS Chen et al [35] in Spiral-Multi-Path QoS Routing presented some bandwidth
reservation from source to destination exploiting the concept of multi paths. But this
model heavily relies on MAC layer too. This is basically spiral path and multi path
combination and a new protocol is formed. Here bandwidth requirement is split into
43
some smaller bandwidth requirements and multiple paths are used to carry the traffic.
This adds extra processing and more calculation is involved in, which is very difficult
to manage in dynamic environment of mobile ad hoc networks.
Bandwidth reservation is applied on multi paths, so that the paths are reserved for
particular sessions, this is very hard in dynamic ad hoc networks where resource
guarantee is very difficult to manage. The division in sub paths even posses some
serious drawbacks as now guarantee is required on more paths. This type of
reservation is hard to support in dynamic mobile ad hoc network.
OLSR is also extended to support Quality of Service [36]. It was claimed that OLSR
suits well in dense and large networks. Neighbor detection is done through same old
HELLO messages. Multi Point Relaying is the specialty of OLSR. Maximum
Bandwidth and Minimum Delay are taken as metrics. Here each node is responsible
for keeping record of QoS status with all its neighbors. MPR selection is dependent
upon required QoS level. Shortest and widest path entries are there in routing tables
maintained by each node. These entries take into account the required maximum
bandwidth and minimum delay metrics.
Table 3.2 Summarizes some of the extended routing protocols for Quality of Service
provisioning in Mobile Ad Hoc Networks. It is clear that these are not mature
solutions and none have compared its performance with others in this category.
Further these are based on conventional approaches hence are purely deterministic in
Routing.
44
Table: 3.2: Extended MANET Routing Protocols for QoS
S. No.
QoS Extension applied to
Routing Strategy
Routing Scheme
QoS Metric
Addressed
Parameter analyzed
Compared with other
Mobility/Lo
ad Supported
1 DSR Deterministic Reactive Delay Packet Drops
No No
2 OLSR Deterministic Proactive Bandwidth & Delay
Packet Drops No No
3 AODV Deterministic Reactive Bandwidth & Delay Throughput No No
3.4 Conclusion In this chapter routing in Mobile Ad Hoc Networks is discussed. A literature survey
about MANET Routing Protocols is presented. Some leading proposed protocols are
discussed in detail. QoS aware routing is explored and some best effort routing
protocols extensions are also discussed.
It should be noted that all the above mentioned protocols having QoS extensions rely
on feedback of lower layers and also have some other potential drawbacks. Therefore
the above mentioned techniques for QoS provisioning in Mobile Ad Hoc networks can
not yet give a mature single solution due to special described characteristics of such
networks.
45
In the next chapter, ineffectiveness of conventional routing approaches is
demonstrated by performing the simulation of two leading MANET routing protocols
and the result obtained are analyzed accordingly.
46
CHAPTER 4
DSDV AND DSR SIMULATION
4.1 General There exist a lot of routing protocols proposed for Mobile Ad Hoc Network
environment so far. It was described in earlier chapters that most of these routing
protocols belong to two main streams, namely Proactive routing protocols and
Reactive routing protocols. Before proposing any solution, it is essential to understand
the working mechanism of conventional approaches and investigate that which routing
protocol gives better performance under traffic load and node mobility conditions
which are the basic routing functionality parameters of MANET. For this purpose, two
popular protocols namely Destination Sequenced Distance Vector (DSDV) and
Dynamic Source Routing (DSR) have been thoroughly analyzed to observe their
performance. A number of scenarios have been created to observe their performance
by simulating both the protocols on NS2 simulator. The results of the analysis of these
two protocols show inconsistency and ineffectiveness even handling mobility and
load. Considering Bandwidth as the main parameter, it is missing in these two
simulated protocols.
For developing better understanding of the subject, performance evaluation of two
leading MANET Routing Protocols namely Dynamic Source Routing and Destination
Sequenced Distance Vector has been carried out by simulating both the protocols on
NS2 simulator [24]. These evaluations were carried out by generating different
47
scenarios. Different loads and mobility patterns were introduced in these generated
scenarios. Results showed weakness and characteristics of both the protocols. Detailed
simulation and results are presented and discussed in this Chapter. Different protocols
hold diverse characteristics and behaviors to be adopted in different possible
situations. In Ad hoc network, mobility and traffic are the two main factors which can
vary frequently and influence the behavior of the network as a whole. Therefore here
simulation of DSR and DSDV routing protocols has been carried out on varying
mobility scenarios and at different traffic patterns in a number of experiments. The
analysis has been carried out on NS2 simulator.
4.2 Simulation Environment All the simulation work has been carried out on Intel 1.4 GHz Linux machine having
Fedora 6.0 operating system installed and running. The network simulator NS2
version 2.29 is used for simulating Adhoc routing protocols. The languages used for
writing the simulation script was Otcl and well known C++ language. Beside these,
other tools which proved helpful during simulation and trace file analysis are XGraph,
jTrans and Matlab version 7.4. Table 4.1 shows the summary of these parameters used
in carrying out all the simulation experiments.
48
Table 4.1. Basic Factors
Factor Value
Processor 1.5 GHz
Operating System Linux RedHat
Simulation Environment Ns 2.29
Programming Language C++ and Otcl
Extra Matlab 7.4, Xgraph, jTrans
4.3 Simulation Overview A general simulation methodology by using network simulator is given through
flow chart in Figure 4.1, which shows three major phases of NS-2 simulation
including phase before the simulation, phase during the execution and phase after the
simulation. All the phases are discussed in the following subsection.
49
Figure 4.1. Simulation of Three Phases
EXECUTION
Scenario Generation
Scenario file Traffic Pattern
Otcl Script writing
Trace file NAM file
Network Simulator
Data Processing
jTrans Awk/perl
Matlab XGraph
PRE-SIMULATI
POST-SIMULAT
50
4.3.1 Pre-Simulation Phase
Pre-simulation phase consists of the process of generating the scenario file which
explains the number of nodes, topology, model of movement of the nodes and etc.
This includes the generation of communication file which explains the traffic model in
the network, the final step is to write Otcl script for the simulation of the particular
protocol.
4.3.2 Ns-2 Execution Phase Execution phase gets input through the Otcl script written in preceding phase and
generates a rough form of data called trace file. This file contains the results obtained
through simulation execution and gives the complete information of time of execution;
participating nodes; how much data and control packets arrived; transmitted,
forwarded and dropped.; and etc
4.3.3 Post-Simulation phase
The major goal of the post-simulation procedure is to extract the useful information
from a rough and huge data file which is the previously discussed trace file generated
in execution phase. Scripting languages like awk and perl can be utilized to convert
the trace file in that format which is suitable for XGraph, but an alternative method is
to use application like jTrans to understand the trace file and then use matlab for
interpreting graphically.
51
4.4 Simulation Setup We have used 50 nodes with geography of 500x500 meter square and the traffic
application used is of type constant bit rate with a maximum data rate of 8 packet/sec
as shown in Table 4.2.
Table 4.2. Simulation Setup Table
Parameter Values
No. of nodes 50
Environment size 500x500 m2
Simulation time 200 sec
Traffic type CBR
Data rate 8 packets/sec
Propagation Model Two ray ground
Mobility speed[max] 5 m/sec
Antenna type Omni
Antenna height 1.5 m
4.4.1 DSDV Simulation
We have simulated the Destination Sequence Distance Vector (DSDV) routing
protocol by using above simulation setup, and the results obtained through trace file
are shown in Table 4.3.
52
Table 4.3. DSDV Simulation Results
Property Name Value
Simulation length in seconds: 199.61050957999998
Nodes: 50
Sending nodes: 50
Receiving nodes: 7
Dropping nodes: 48
Packets generated: 2962
Packets sent: 2849
Packets received: 3488
Packets forwarded: 100
Packets dropped: 113
Minimum generated size of packet: 32
Maximum generated size of packet: 620
Average generated size of packet: 172.0
Generated Bytes: 512014
Sent Bytes: 487094
Received Bytes: 522206
Forwarded Bytes: 53200
Drop Bytes: 24920
There are different types of packets which can be classified in two broad categories
namely data packet and control packets. Data packets refers to the actual constant bit
rate (CBR) traffic where as control packets include Clear to Send (CTS) packets,
Request to Send (RTS) packets, Acknowledgement packets (ACK), Address
Resolution packets and number of packets specific to the routing protocol for route
finding.
53
Figure 4.2. Traffic Bytes Vs Time
Figure 4.2 shows that initially only DSDV message control packets are sent to find the
route because DSDV being a proactive routing protocol, therefore uses table driven
approach for route finding in which routing table already contain routing path entries.
Here in real time system, minimum end to end delay is of prime interest. Ratio
between control packets and actual data is important here. The control information is
represented by red color while the actual data is represented by blue color. So as in the
54
case of table driven approach the redundant information is a lot as compared to actual
data.
Figure 4.3. Packet ID Vs End to End Delay
Figure 4.3 shows the end to end delay encountered by each packet which includes the
delay at MAC layer, network layer and application layer. It is clear from the figure
that most of the packets encounter a smaller delay of 0.005 seconds and some have in
the range of 0.01 sec to 0.03 sec while a few packets have delay higher than 0.03 sec.
It is obvious that as in the case of table driven approach the initial delay is not much
55
because the routes are already known but this would not be the case in DSR where the
approach is on demand and there should be more initial delay.
Figure 4.4 depicts the number of each control message and data message for given
simulation setup of destination sequenced distance vector routing protocol. This is
basically a Pie chart showing number of packets for actual data and control
information. ACK is the acknowledgment packets at MAC layer. RTS and CTS are
MAC layer Request to send and Clear to send packets.
Figure 4.4. Number of packets
56
4.4.2 DSR Simulation
We have simulated the Dynamic Source (DSR) routing protocols on same parameters
as used for previous DSDV protocol, and got the results shown in the Table 4.4.
Table 4.4. DSR Simulation Results
Property Name Value
Simulation length in seconds: 170.968715759
Nodes: 50
Packets generated: 1736
Packets sent: 1720
Received packets: 2229
Forwarded packets: 114
Dropped packets: 16
Minimum generated size of packet: 32
Maximum generated size of packet: 556
Average generated size of packet: 109.0
Generated Bytes: 189538
Sent Bytes: 187938
Received Bytes: 199474
Forwarded Bytes: 53924
Drop Bytes: 1600
Unlike previous destination sequenced distance vector (DSDV) the dynamic source
routing (DSR) protocol is a reactive routing protocol so route finding process starts
57
only when data traffic arrives. The CBR data will be same like DSDV as shown in
Figure 4.5. Here the ration between overhead and actual data is much improved
against previous case where control information is very much as route searching
process goes on without need.
Figure 4.5. Traffic Bytes Vs Time
It is evident from Figure 4.5 that DSR has less number of control messages. If there
are less number of control messages, then less byte overhead will be involved making
optimal use of available bandwidth which will lead to less packet drop ratio.
58
Figure 4.6. Packet ID Vs End to End Delay
In comparison with DSDV from Figure 4.6, the DSR protocol show higher end to end
delay. About 98% packets encounter a delay up to 0.025 sec and some of packets have
even more delay just because these packets have to wait in buffer until RREQ finds
the optimal path to send the data.
59
Figure 4.7. Number of packets
There is significant difference between the routing overheads of DSDV and DSR
protocol. The number of routing messages in DSDV were 1461 which make the
overall byte overhead of 1461*32=46752 bytes, whereas looking at Figure 4.7, DSR
has comparatively very less number of routing control packets of 41 which make the
byte overhead of 41*32=73 bytes. Here the message is represented with CBR traffic
which is 232, so a significant ratio achieved between actual message and control
overhead.
60
4.5 DSR and DSDV Result Analysis In this Section we will analyze the behavior of previously simulated reactive and
proactive routing protocols at varying mobility and traffic load conditions. There are
several issues in routing in MANET; this is the reason why the proposed protocols are
not yet properly implemented. Like if one improves one aspect then the other related
issues are loosely considered, as a result the output is not up to the acceptable desired
level. There is no any single scheme which fits good for all situations. Here our main
focus is to analyze that which of our main leading protocols perform better in what
situations.
4.6 Measures of Interest Certain parameters are considered in Network Simulation like how much packets
received at the end and how much dropped. Overhead in networking is also taken into
account for comparisons. Therefore we will consider three important parameters, at
our output as a base for our comparison which includes:
Data throughput
Protocol byte overhead
Packet drop ratio
These are explained in the following sections.
61
4.6.1 Data throughput
The ratio of data throughput can be defined as average ratio of the successful packets
delivered to particular destination to those of generated by the traffic sources. This is
measured as bits per second.
4.6.2 Routing Overhead Data packets transmission, there are routing packets that are transmitted for
maintaining links and updating routing tables which is treated as control information
and poses routing overhead for each protocol.
4.6.3 Packet Drop Ratio
Packet Drop Ratio is the measure of determining how many packets fail to reach at the
destination and are dropped. This is against the throughput and is measured in number
of packets. Our interest is to minimize this ratio at maximum possible level.
4.7 Mobility Analysis Mobility is relative movement of the nodes with respect to others. Mobility of the
scenario can vary by changing the two important factors namely Node Speed and
Pause Time. The simulation setup for this analysis is shown in Table 4.5 with respect
to different required parameters.
62
Table 4.5. Parameter Table
Parameter Values
No: of nodes 50
Pause time 1.0 sec
Environment size 500x500 m2
Simulation time 200 sec
Traffic type CBR
Data rate 2 packets/sec
Propagation Model Two ray ground
Mobility speed[max] 20 m/sec
Antenna type Omni
Antenna height 1.5 m
Node speed corresponds to the actual speed of the participating node. Here if its value
is zero then it means that the node is stationary and therefore behaves like a fixed
network and max speed up to 20 m/s corresponding to the speed of the vehicle. Pause
time is also one of important factor which affects the mobility. If a scenario has pause
time then it indicates the high mobility and vise versa.
4.7.1 Throughput Vs Mobility Figure 4.8 shows throughput at the varying mobility scenarios of nodes within the
network.
63
Throughput analysis
0.88
0.9
0.92
0.94
0.96
0.98
1
2 4 6 8 10 12 14 16 18 20Mobility Speed (m/sec)
Thro
ughp
ut
DSRDSDV
Figure 4.8. Throughput Vs Mobility
Here it is evident from the graph that the DSR protocol shows a 2% variation in the
throughput when the speed of vehicle is 20 m/sec. In case of DSDV, as the mobility
increases, the throughput decreases rapidly correspondingly. It means that at high
mobility, the DSDV has unacceptable results, and even for low mobility or zero
mobility, speed throughput is not 100%.
The reason is that the DSDV being proactive routing protocol assumes that all the
routes are available before hand. These routing tables first start building up when the
64
network appears and converge with the passage of some time. Initially when
convergence is yet to occur but if communication is started then some initial packets
must have to be dropped because all the paths are not known at this stage.
4.7.2 Protocol Overhead Vs Mobility
Protocol overhead is an important factor to be analyzed because more control packets
mean more wastage of resources like bandwidth and power. The value we have plotted
does not look into the MAC layer overhead it counts only the overhead at network
level. Figure 4.9 depicts the corresponding overheads of both DSR and DSDV
protocols.
Overhead analysis
1
2
3
4
5
6
7
8
9
10
11
2 4 6 8 10 12 14 16 18 20Mobility Speed (m/sec)
Ove
rhea
d
DSRDSDV
Figure 4.9. Protocol Overhead Vs Mobility
65
It is clear from Figure 4.9 that the byte overhead of DSDV is fairly constant even in
the scenarios when mobility tends to be at higher side. This is due to the proactive
nature of DSDV which updates routing information by periodic broadcasts. In case of
DSR one may observe some sort of strange behavior as byte overhead increases
rapidly at some stage. This is because of the increased mobility, which in turn will
have more rapid topology changes so more control information will be sent to find the
routes. This is mainly because the DSR is on demand based, therefore route requests
are broadcasted only when there is a need for route exploration and the complete
routing path is appended in the header of data packet thus increasing the byte overhead
respectively.
4.7.3 Packet Drop Ratio Vs Mobility
Packet drop ratio gives a clue to know how good that protocol is to be when used .The
graph in Figure 4.10 shows the variation between the packet drop ratios at different
mobility speeds. It is evident from aforementioned study that DSR enjoys up to 2%
drops in packets only even at very high mobility speed of vehicle [20 m/sec, this
corresponds to 72 Km/hour which is the average standard speed of a moving vehicle
and very near to real life scenarios]. In case of DSDV about 10-12% packets drop with
the increase of mobility speed.
Figure 4.10 shows that though DSDV has constant protocol overhead but it is
inappropriate for the situation where the environment is of high mobility, definitely
due to its proactive nature.
66
Drop Ratio analysis
0
0.02
0.04
0.06
0.08
0.1
0.12
2 4 6 8 10 12 14 16 18 20Mobility Speed (m/sec)
Dro
p Ra
tio
DSRDSDV
Figure 4.10. Packet Drop Ratio Vs Mobility
4.8 Load Analysis Traffic load is another important parameter to be investigated and thorough analysis
has been conducted too in this regard. We have mainly three factors by which offered
load can be adjusted. These include data rate, number of CBR flows and the Packet
size.
67
From all these three factors, data rate is the most dominant one which influences the
load more than others do. We increase the traffic load by varying the data rate in our
simulation analysis. Table 4.6 summarizes the parameters used in our simulation
experiments of Load Analysis.
Table 4.6. Load Analysis Parameter
Parameter Values
No: of nodes 50
Environment size 500x500 m2
Simulation time 200 sec
Traffic type CBR
Data rate 2,4,8,10 packets/sec
Propagation Model Two ray ground
Mobility speed[max] 5 m/sec
Antenna type Omni
Antenna height 1.5 m
4.8.1 Throughput Vs Load
It is found that the significant change in throughput occurs when data rate is varied
accordingly. It is clear from the Figure 4.11 that the DSR performed best when the
data rate is around 5-10 kbps. But when the load increases beyond the range of 18kbps
the DSDV routing protocol performs better than the DSR routing protocol.
68
Throughput analysis
0.4
0.5
0.6
0.7
0.8
0.9
1
0 5 10 15 20 25 30 35 40 45Data Rate (Kbps)
Thro
ughp
ut
DSRDSDV
Figure 4.11. Throughput Vs Offered Load
On the whole as the data rate increases, the throughput by both protocols decreases
with same rate.
4.8.2 Protocol Overhead Vs Load
At low data rate, the DSR protocol offers less overhead compared to DSDV protocol
but when data rate increases beyond 15 kbps it is the DSDV which offers less
overhead.
69
Overhead analysis
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 5 10 15 20 25 30 35 40 45Data Rate (Kbps)
Ove
rhea
d DSRDSDV
Figure 4.12. Protocol Overhead Vs Offered Load
On the whole the protocol overheads of DSR protocol and DSDV protocol are
comparable to each other and remain constant over increasing data rates as is evident
from Figure 4.12.
4.8.3 Packet Drop Ratio Vs Load
The graph below shows that how number of packet drop ratio varies when data rate is
increased. Both DSDV and DSR packet loss increase with increase in the load.
70
Packet Drop analysis
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 5 10 15 20 25 30 35 40 45Data Rate (Kbps)
Dro
p R
atio
DSRDSDV
Figure 4.13. Packet Drop Ratio Vs Offered Load
The increased send rate also set demand on the send buffer of the routing protocol.
Congestion occurs and packets are dropped. As can be observed from Figure 4.13, if
data rate is below 18 kbps, the DSR has less packet loss, but as data rate increases
from 20 kbps, DSDV offer slightly better performance.
4.9 Conclusion In this chapter two MANET routing protocols are simulated on popular NS2 platform.
Simulation process is presented in detail. Major phases are described along with detail
71
of environment parameters. Results are elaborated to show the fractions of control
packet and their ratio with actual data packets. End to end delay is summarized and
analyzed as comparison between both the protocols.
A comparison analysis is performed between two simulated MANET routing
protocols. The objective is to determine which protocol can be considered as single
protocol for all possible scenarios. The analysis was performed by varying mobility
and load. It is found that DSR is preferable over DSDV in high mobility and low load
[below 15 Kbps] condition even though DSDV offer constant byte overhead in
varying mobility and load. On the other hand DSDV has slightly better performance
in high load conditions.
On the basis of this simulation it can be concluded that conventional approaches are
not that much capable to give a single mature solution for routing needs in Mobile Ad
Hoc Networks.
In light of the analysis performed on representative protocols, there are two options to
deal such cases: (i) to introduce innovation in the existing protocols so that they can
perform in a way that can bring QoS feature in it, (ii) to use an independent approach
that can find the routes based on bandwidth available on the given routes.
We have used the second option because: (i) it will not depend on the existing
protocols features (ii) Due to the adaptive feature of our proposed approach, the
proposed solution will classify routes from the existing network in graded fashion
72
emerged from the existing conditions of load and mobility. The subsequent chapters
will elaborate the details of proposed approach used for QoS aware routing protocols.
73
CHAPTER 5
SWARM INTELLIGENCE
5.1 General A long ago it was considered a myth that how natural species like birds adapt to
collective systems for achieving larger objectives with very less direct interactions.
Nature has inspired researchers to ponder upon different nature phenomenon to sort
out solutions for problems on the pattern of these natural solutions. This interest took
the research community in biological world and different studies were carried out on
species like ants, birds, undersea living beings too. It was found that there exists some
intelligence to manage social behaviors of insects and birds. This is called Swarm
Intelligence. E. Bonabeau et. Al. [37] defines Swarm Intelligence (SI) as the
characteristic of a system in which the combined performance of unrefined agents
does local interaction with their surrounding cause rational practical universal model
to appear.
It was revealed that these natural species follow some interesting patterns in their
livings. Social cooperation is a beautiful example among other characteristics found in
such biological agents. Emerging swarm intelligence from these systems gives
interesting phenomenon. Following are the basic feature of swarm intelligence
systems:
Interaction between agents is distributed and local
74
This local interaction yields global behavior
These agents are usually very large in number
Communication between these agents is indirect
Above communication is different in different systems
One can see birds on the sky following some pattern while traveling even in long
distances. Deneubourg et.al. [38] presented the phenomenon of termite digging.
Stigmergy is used to encourage more digging by termite. In start the termite moves in
random but later on moves probabilistically. While extracting a soil particle the
pheromone is laid down. Pheromone concentration increases with the passage of time
which attracts more diggers.
5.2 Advantages of Swarm Intelligence From the study, it is found that concerted effort is made by thousands of ants in search
of food, migration or any other activity. Ants exhibit this formation in a lovely manner
by living in colony like formation. Ants remained very common and near to mankind
since almost the birth of mankind. Ants collective behavior achieve very complex and
large solutions. N. R. Franks [39] gives the following examples of such ant behaviors.
(i) Collective movements
(ii) Combined efforts
(iii) Living in colonies
(iv) Collective food search
(v) Traversing the shortest path
75
(vi) Cooperation in complex tasks
a. Dragging large foods
b. Forming bridges
These properties of biological agents tempted researchers to study such behaviors in
larger interest of adapting these in problem solving strategies.
From the outcome of above studies, new techniques emerged for solving optimization
problems [40, 41]. These techniques got inspiration from emergent swarm intelligence
[3, 42] behavior of social insect colonies.
Usually thousands of small agents like ants with very little direct communication solve
global optimization problems like shortest path to a food source without any central
supervisory role [43].
Ants employ a chemical matter named pheromone in order to communicate indirectly
among them. This pheromone is laid down by the ants while traveling in a path.
Consequently this pheromone is sensed and rests of the ants follow the same path as
long as pheromone remains available. Pheromone concentration increases as more ants
traverse the same path which attracts other ants too, hence convergence appears.
Diffusion property of this pheromone is an important one for mobile adhoc networks,
due to which pheromone decays as far as that path is not used, hence stale routes are
omitted from routing tables which is a great requirement of such dynamic ad hoc
76
networks. The process of indirect communication between agents by modification
induced in environment through this pheromone is known as stigmergy.
5.3 SI based Routing Research in swarm intelligence has revealed great matching properties between this
stigmergy and communication networks routing. Therefore some algorithms have
been proposed for wired networks. Most popular of these is Antnet proposed by
Gianni Di Caron ad Marco Dorigo. Getting inspirations from this popular Antnet a lot
of research gave birth to more routing protocols [44] like this Antnet.
Quality of Service provisioning is also touched upon by swarm intelligence
researchers and some of the protocols have been proposed [45] based on this ant
colony.
The basic differentiation between swarm intelligence routing and traditional routing
algorithms is that the former uses stochastic exploration of routes depending upon the
pheromone concentration.
In a simple routing, ants are used as mobile agents for route exploration much like ant
foragers in real ant colonies not having the global picture of whole of the traveling
path [46]. There are usually two types of ants, one is forward and the other is
backward ant. Forward ants are sent for exploration and after reaching at destinations
these forward ants die out and transfer acquired network information to backward ants,
which in turn are usually source routed back to source while updating the routing
77
tables at intermediate nodes. Routing tables contain updated pheromone values as a
measure of good put of using the corresponding path.
Swarm intelligence based routing protocols are robust, dynamic and rely on local
information only. Further multiple routes are also achieved through ant colony based
algorithms. These are also the desired properties of MANET routing protocols. These
protocols have proven results for lesser overhead and more efficiency.
Gianni Di Caro and Marco Dorigo are considered as the pioneers of Ant colony
optimization technique. They proposed the first routing algorithm employing the ant
colony for communication networks called AntNet [47]. In AntNet, every simulated
ant makes a trail beginning at its source and ending at its destination. Whilst making
that path, this ant gathers the detail about the duration of the path factors and implied
detail about the load condition within the network. The same detail is afterwards
reversely transferred through another ant traveling in the reverse track and is utilized
to alter the entries in routing tables of the nodes which have been passed by.
The authors compared it with the well known shortest path and distance vector routing
algorithms and through simulation, its best performance was confirmed. The authors
claim also that AntNet has merely a little set of adjustable parameters and its
performance is much stronger with respect to their adjustment.
Ant Colony Optimization is beautifully applied to traveling salesman problem [48,
49]. Here the objective is to connect n given cities with minimum lengths so that every
78
city is only once visited. Now assume ijd be the distance between cities i and j. This
scenario can graphically be represented by considering nodes as N and E are the edges
of graph denoting the distance between the cities.
Now applying ant algorithm, an ant traverse from city to city to complete the trip. At
every cycle of the algorithm, each ant k, k=1,...,m, traverse a trip by taking n = || N ||
steps; note that here the decision is probabilistic in nature. Each cycle is indexed by t,
max1 tt where maxt is the maximum number of permissible trips.
Here every ant traverse from city i to city j while depending on:
To see if the said city has been prior visited or not. There is in built memory to
every ant. The said memory increases as trip goes on and is purged out as the
trip finishes. It is the memory which identifies a set kiJ of those cities which
the ant still needs to visit when it reaches city i. While at the start kiJ includes
all cities except the originating city. Memory helps discarding the city which
is visited already.
The opposite of the distance isij
ij d1
. This is known as Visibility. Its utility
is directing the search operation of ant. This Visibility is dependent on entirely
the local information and denotes the attractiveness of selecting city j while in
city i.
Quantity of implicit pheromone trail ij over the link joining city i to city j.
This Pheromone trail helps representing the previously achieved desirability
79
of selecting city j while in city i. The said Pheromone trail is on-line updated.
Hence the global information emerges.
Here golden rule named random proportional transition is presented which is the
probability for ant k to travel to city j from city i when making its tht trip. This is
reflected by following equation:
)]([)]([)]([)]([
)(ttJl
tttP
ililki
ijijkij
(5.1)
Here α and β are the 2 parameters which are adjustable and whose function is to be in
charge of the relative weight of trail of pheromone )(, tji and visibility ji , . It is
worthwhile to note that when α approaches to 0, then the selection would go to cities
which are most close. Now when if β approaches to 0, then the quick reinforcement is
applied and consequently optimal routes are obtained.
Now the performance quality of every ant k derives an amount of pheromone )(tkij
is laid down by every ant k after completion of a trip on each edge (i, j) which it has
used.
This is shown in the following equations.
),(),()(/)( tTjiiftLQt kkkij (5.2)
otherwisetkij ,0)( (5.3)
80
Here )(tk is the completed trip of ant k at t iteration, )(tLk is its length, and Q is a
parameter which is same in order of magnitude of the trip length.
Pheromone evaporation is very beneficial phenomenon. The pheromone should
undergo a decay, otherwise all the ants will eventually follow the same route so that
the solution turns from optimal to none one. Pheromone decay is incorporated by
defining a parameter of decay ρ, 10 . As a result, the rule for pheromone update
is then as following equation:
)()().1()( ttt ijijij (5.4)
Here )()( 1 tt kij
mkij , and m denotes ants. Look that initially pheromone is in a
small value 0 .
Above all, is a manifestation that number of ants plays a significant role.
Popularity and promising nature of AntNet inspired researchers to propose solution
suggesting suitable improvements in this basic version of algorithm [50].
Dynamic topology of MANET and special characteristics due to wireless links etc.
demand efficient and robust routing protocols for such type of communication
networks, rendering conventional routing strategies inefficient. Swarm intelligence
particularly ant colony based algorithms are tailored for MANET needs and showed
good results [51, 52, 53].
81
G. Di Caro in ANTNET-FA [54] suggested the following improvements to the basic
ANTNET algorithm for MANET. Forward ants and backward ants utilize of high
priority buffers. The backward ants also renew the entries in route tables of the nodes
which are visited utilizing estimation of the local traversing duration. This is in
contrast to the time which was taken by corresponding forward ant.
The lessening dynamics of the link queues is modeled in terms of an identical and
invariable process D that is dependent only on the bandwidth lb of that link. This
means that exhaustion of the queue is not slowed down by both the sender’s process
and the consuming process present at the other side of that link. These suppositions are
suffice to say that the transmission of the packets which are coming up at the link
queue klL to the connected neighbor is concluded following an interval of time
l
lllll
kl b
qddbqD );;( (5.5)
Here lq is the total bits enclosed in the packets those were waiting in the queue klL ,
whilst lb and ld are correspondingly the bandwidth of the link and delay of the
propagation.
In accordance with the said lessening model, the estimation of time lkT which would
be taken by a packet of data of the same length as of the forward ant to travel to the
neighbor l from which the backward ant originated is calculated as:
82
),;( llalkllk dbsqDT (5.6)
The value of mkT of the node m on the ant path V is calculated by summing up the
time for every hop:
mkvvvviivvvi
mk VTTmmkk
kmm
},,...,{,1},...,,{
}11,
11
(5.7)
This is in comparison with the last model, where ants were moving in lesser pace
while the available paths remained often packed. Here the forward ants try their level
best to traverse the relevant path as soon as possible to get to the destination. Now
while doing so, forward ants seems flying instead of walking and due to this particular
reason, this new algorithm is known as AntNet-FA, so that the abbreviation FA is
meant for ants which seems flying. Authors claim the efficiency of ANTNET-FA
better than ANTNET even in case of more number of nodes.
AntHocNet [55, 56, 57] is another popular ANTNET based routing algorithm tailored
for MANET needs. This is hybrid in nature as it uses reactive route exploration while
after initial route discovery its proactive component constantly searches alternate route
which helps in case of previous link failures.
5.4 Conclusion In this chapter, Swarm Intelligence is discussed in detail. The basic Antnet protocol is
explored to determine the effectiveness of Swarm Intelligence based techniques.
Examples from real life scenarios were discussed. Applicability of these Swarm
Intelligence based approaches in MANET is also explored. Stigmergy of ants is
83
elaborated to see its strength in solving routing problems in dynamic ad hoc
environment of MANET. Recognizing the great potentials of swarm intelligence based
solutions, a novel routing algorithm is proposed in this thesis for supporting quality of
service in MANET. This is discussed in the following Chapters.
84
CHAPTER 6
PROPOSED SOLUTION
6.1 Introduction Popularity of Mobile Ad Hoc Networks demands quality of service provisioning in
addition to best effort services. Swarm Intelligence based effective and promising
solutions have inspired researchers to propose QoS aware routing solutions in this
domain.
In chapter four, performance evaluation was performed on two MANET leading
routing protocols based on conventional approaches and results demonstrated the
ineffectiveness of both the protocols in terms of basic functionalities such as load and
mobility. This revealed the incapability of conventional approaches and addressing
QoS appeared more challenging for these conventional approaches. Our proposed
solution uses new technique based on the principal of Swarm Intelligence to address
the QoS requirements in MANET. The bandwidth metric is supported in the proposed
algorithm which the conventional approaches are incapable of handling. The
conventional approaches showed ineffectiveness even handling mobility and load,
hence bandwidth requirement is even more challenging for those to handle, therefore
approach of Swarm Intelligence is used and a basic version for Traffic Classification
has been proposed considering the Bandwidth as the main parameter which was
missing in previous simulated protocols.
85
6.1.1 Basic Concept Behind the Proposed Approach
In any given network, there are two types of flows in general: BE flows which
requires the data to be delivered to the destination, and QoS flows which apart from
just delivery reliability, requires some additional constraints such as available
bandwidth to be satisfied.
Instead of transmitting entire data through one route, the proposed algorithm discovers
during transmission new efficient paths that are bandwidth specific. The proposed
Swarm Intelligence based algorithm is of adaptive nature and finds BW specific routes
because the proposed protocol does not only have the capacity to give BE (Best
Effort) routes (which is the routine feature of conventional routing) but it also finds
other routes graded for different class of service to cater for different applications
requiring different BWs.
Now, in this case, service classes are made available through the proposed solution in
order to serve according to priority of the traffic. The solution uses an adaptive
approach giving alternate QoS supporting paths, which are instantly made available in
case of degraded performance that often happens in dynamic environment of MANET.
In our case, the most important parameter of QoS is considered in terms of Bandwidth.
Our SI based protocol defines traffic paths in the network of various classes belonging
to different range of bandwidths. Thus our solution will be independent of load and
mobility constraint and with in the available network , it will identify the most suitable
paths of graded traffic where information signals can be diverted to these paths based
on their bandwidth requirements.
86
A new QoS aware Routing Protocol for MANET called Ants inspired Traffic
Classification in Mobile Ad Hoc Networks is proposed here. Stigmergy in ant colony
[58, 59] is the core of this algorithm, which is a process of indirect communication
between agents through modifications induced in the environment. This idea is
basically derived from ants which adapt the shortest path with the passage of time.
They do it through stigmergy process.
Famous example in this regard is of food searching in ant species. Ants search food
and carry this to their nest in a very interesting way. Though individual ants do not
posses much intelligence but their collective collaboration solves complex shortest
route problem. Ants utilize a chemical substance known as pheromone for indirect
communication between them. This phenomenon is called the stigmergy, which is
indirect communication between agents through modifications induced in
environment. Pheromone concentration increases as more ants traverse the same path
and leave their own pheromone trail subsequently which attracts more other ants.
Figure 6.1 shows how ants start their food searching journey and adapt to shortest path
finally.
87
a) Ants traversing a straight path
b) Ants having an obstacle along the path
c) Ants traversing the shortest path
Figure 6.1. Ants flow between source and destination
In Figure 6.1 (a), ants start traversing from source to destination and then back simply
along straight path while depositing pheromones along the path they travel. In Fig. 6.1
(b), there comes an obstacle along the path, here initially some ants choose shorter
88
path while some go along longer path, choices of path is dependent upon nothing at
this stage but random one. With the passage of time concentration of pheromone
increases along shorter path than longer path as more ants are attracted to higher
density of pheromone. Therefore ultimately ants adapt to this shorter path. Fig. 6.1 (c)
is the resultant shortest path adaptation from ants.
The above phenomenon is similarly mapped to communication networks for
optimization/ convergence and other issues too. This proposed solution is heavily
dependent on natural ants like strategies.
Next hop decision during the course of transmission is of course probabilistic
one. Following describes this well. Take dstDAll as all the next hop neighbors to the
required destination dstX . Therefore a packet reaching at node a will be sent to node b
with following probability. Equation 6.1 shows relation between route length and
quantity of pheromone in simple ant colony optimization [60].
)(1)(
tLt s
sji (6.1)
Here )(tLs is the route length, )(tji is the concentration of pheromone at t
time at jth node where next hop is node i.
6.2 The Proposed Model In this section, the novel QoS aware routing protocol is proposed which is Ants
Inspired Traffic Classification in Mobile Ad Hoc Networks.
89
Recalling the previously discussed two main strategies for route finding in a MANET,
our protocol is initially like on demand in nature. Route exploration is only carried out
when required, so that unnecessary control traffic is minimized. However, once data
path is established, other paths to the same destination are explored so that in case of
link failure, data could be forwarded in acceptable small interruptions. This is greatly
helpful in case of QoS applications. The algorithm assumes bidirectional symmetric
links all the time. Initially links to neighboring nodes discovery is carried out through
periodic beacon messages.
Ant colony gives impressive concept of pheromone diffusion. This is of great help in
our approach too. According to this approach, hello messages (that neighbor’s send to
each other for ensuring their presence) contain information about active destinations.
These are basically those nodes which have recently participated in a communication
session. The active destinations are defined to be those nodes with which currently
communication had taken place. As a result all the neighbors now about their one hop
neighbors accordingly
Swarm intelligence has built-in property of negative reinforcement, which gives birth
to pheromone evaporation. Here backward ant will also update a path that has lost
quality. Moreover, we will use pheromone evaporation phenomenon of ant colony
according to which goodness of a route present in the routing table decreases with
time without even checking that route. So in this way we will remove those entries
90
from the routing table which are not in use by any application thereby, keeping the
routing table as short as possible.
Following is the equation which exponentially exploits evaporation property of
pheromone for removing old unused routes from route caches.
)]()()[1()1( ttt jijiji (6.2)
Here ρ is the constant of pheromone evaporation where 10 , and )(tji is the
pheromone increase. It is obvious that smaller the quantity of ρ, evaporation of
pheromone is slow. Further as pheromone concentration increases the evaporation
process is rapid.
Only two types of ants as agents are used, forward ant and backward ant. Whenever
there is a need for data transmission, forward ants with required QoS level are
broadcast to the desired destination from intending source. Backward ants on the other
hand are unicast in nature towards destination (originally the source). Forwards ants
keep record of visiting nodes in a stack. On reaching the destination, these die out and
give birth to the corresponding backward ant by transferring the stack too. This
transferring may also be done on an intermediate node, which has a routing table entry
matching the desired QoS metric to that particular required destination. Backward ants
gather information from Forward ants and based on this fresh information backed by
backward unicast path, modify the routing table entries of intermediate nodes.
91
Table 6.1 summarizes the strategies used in delivering all possible type of packets in
the corresponding network.
Table 6.1. Packet delivering strategies
Type of Packet Delivering Strategy
Route Request Packets Broadcast
Route Reply Packets Stochastic unicast
Route Error Packets Deterministic unicast
Data Packets Stochastic unicast
6.3 Traffic Classification Recalling the QoS metrics, Bandwidth is taken as a metric for our QoS aware protocol
because Bandwidth is the primary tool for providing quality of service with respect to
the demanding applications. For example rich applications like Streaming video
require more bandwidth as compared with the normal HTML browsing. It is the
Bandwidth which decides that weather the QoS can be supported in the network or not
and even the latency also relates to the provided bandwidth. Hence the single most
important parameter in QoS provisioning is the Bandwidth.
Four service classes are used as Class A, Class B, Class C and best effort class. Each
packet is marked according to its priority like Class A, Class B or Class C where,
Class A is given the top priority (for multimedia traffic). Class B and Class C are used
92
for traffic with some acceptable losses. If none of the above is found, then it is simply
treated as best effort.
Each class has a lower bound restriction which means that a packet could not be
routed to a link which does not satisfy the lower bound of associated quality of service
class. The bandwidth allocation is done randomly between ranges of 1 Kbps to 400
Kbps. The imposed limits of these defined classes are as follows:
Class A : 71 to 90 kbps
Class B : 51 to 70 kbps
Class C : 2 to 50 kbps
The Best Effort (BE) class could use any link having a bandwidth of 1 Kbps or greater
than 1 Kbps because below this 1 Kbps bandwidth, the link is thought to be lost or
broken and no traffic is routed on that particular link.
Whenever any node generates traffic then the packets are marked according to the
service class of that particular dialogue session. Now at every hop these packets are
served according to the marked priority level until reached at the desired destination
and if that service is not possible then the packets are dropped.
6.4 Routing Information Now here the routing is probabilistic instead of deterministic as in the case of
conventional routing. For example, now if we want to reach at destination d and Pxd is
93
the probability of selecting node x as the next hop, and Pyd is the probability of
selecting node y as next hop then following must be true:
Pxd ← Pixd +(BW)
Pyd ← Pyd – ( (BW)/N-1)
Pxd = 1- ∑Pyd .
Here (BW) is bandwidth used as reinforcement. Only assumption is that: 0
<(BW)<<1, that is bandwidth function decreases proportionally with the available
bandwidth. This is a reflection of the phenomenon that the link having highest
bandwidth availability enjoys the highest probability.
During the data sessions, forward ants are randomly sent to idle paths to explore the
new routes in case of QoS level degradation.
Flow charts are presented for understanding the reactive and proactive route
explorations in the network. In Figure 6.2, forward ant is first broadcast in the mobile
ad hoc network and pheromone concentration is analyzed and if one is found then next
hop decision is made according to the probability value contained in that pheromone
concentration, otherwise rebroadcast of this forward ant is carried out. When the
destination is reached or the node knowing the address of the destination is reached
then this forward ant dies out and gives birth to the backward ant and transfers all the
information to this newly generated backward ant. This backward ant is now unicast to
the destination (the source in this case as the source initiated the request for route) and
94
Start
Broadcast Forward ant
Pheromone Info. present
Choose next hop n with
Broadcast Forward ant
Destination or node knowing dest.
Backward Ant Unicast to Destination
Backward Ant generation
the pheromone concentration is checked and the whole process repeats and hence the
path is setup.
Yes
No
Figure 6.2. Reactive phase of path exploration
95
Start
Create virtual pheromone
If Virtual Pheromone Better
Forward ant dies out
Backward Ant Unicast to Destination
Backward Ant generation
In Figure 6.3, the proactive path exploration carried out during data sessions is shown.
A field named virtual pheromone is created and the backward ant on its way checks if
this virtual pheromone value is better than the previous reactive setup pheromone
value. If this virtual pheromone is better than backward ant it is unicast to the
destination with this better path finding information.
Yes
No
Figure 6.3. Proactive phase of path exploration
The goodness of a path is calculated based on a function of the minimum/bottleneck
bandwidth available. This relation may be modified according to the environments and
96
requirements. Note that outgoing path that could route a packet to any destination is
given values in such a way such that sum of probability values of all those routes
always create one.
The above discussion suffices to describe how routing tables are made. Table 6.2
shows how these Routing tables entries based on probabilities associated with each
node being chosen as next hop are consulted.
Table 6.2. Simple Probabilistic Routing
Destinations
Next Hop
A B C
G 0.83 0.79 0.23
H 0.21 0.91 0.62
J 0.54 0.16 0.41
It is clear from Table 6.2 that whenever there is a packet to be transmitted to a
destination from G, H or J the next hop decision is made on the basis of that particular
probability value. For example if we want to reach at destination H then when we
consult table and the next hop decision is made for either node A, B or C on the basis
97
of probability associated with each. In this case of course node B would be selected as
next hop.
The various fields contained in packet format of both forward ant and backward ant
are shown in Table 6.3.
Table 6.3. Packet Formats of Forward and Backward Ant
Attribute Description
Forward ant Backward ant
Destination
Address
Address of the destination Address of source node
destination Source Address Address of the sending node Address of the destination
Min. BW Minimum Bandwidth
Associated
Minimum Bandwidth
Observed Stack List of visited nodes List of visited nodes
As an example, a scenario can be described for better understanding. Let us consider
Figure 6.4. It is a simple network used for discussion about the approach under
consideration.
98
Figure 6.4. A Sample Network
Node ‘S’ needs a route to destination ‘D’ for a class of service having probability
value 0.3 or above. The forward ant is sent and considering it returns the path through
‘J2’, which is capable to support desired service class so the path is forwarded to
application and communication starts.
In the second phase, proactive forward ants come into action and attempts to find
alternative paths to the same destination. Path through ‘J3’ is rejected because it could
not provide required QoS service class as having the probability of 0.2. Now the path
through ‘J1’ is selected as alternate because it can meet the required probability level
as it is not below to that level.
6.5 Conclusion In this chapter, the ants inspired traffic classification in Mobile Ad Hoc Networks is
proposed. The basic idea used is the stigmergy of ants which is the communication
S
J1
J2
J3
D
0.3
0.2
0.5
99
among different agents through provoking alterations in surroundings. Quality of
service is supported along with basic best effort service by dividing the traffic in three
other classes. These classes are served by the network according to the pre defined
priorities for each. In the next chapter simulation results are presented in support of the
above proposed routing algorithm.
100
CHAPTER 7
RESULTS AND DISCUSSIONS In this chapter, a number of test cases have been described which have been used to
test the proposed Ants inspired traffic classification model for mobile ad hoc
networks.
7.1 Network Topology This is a mesh of 50 randomly placed nodes. Any node can initiate network traffic
arbitrarily. Mobility is highly supported as nodes are very much free to move at their
own and make and break process remain often active in the network. As link breakage
could occur randomly and is provided by allocating zero bandwidth to particular links.
It is also supported to exclude several nodes from test which would cause their
neighbors to have a link of 0 kbps with them, thus no communication is possible in
this particular situation.
Figure 7.1 shows the topology of the network. Here circles represent the randomly
placed nodes and green numbers show the bandwidth of the relevant links. Nodes are
marked by red color in a series though any node can generate the traffic in the network
towards any desired destination. This traffic generation is not time dependent and any
node can send data whenever it needs.
101
Figure 7.1. Network Topology
7.2 Simulations The results are performed mainly in three scenarios and average hop as well as packet
drops are taken as experimental criteria. Firstly in section 7.3.1, 1000 calls are
generated when all of the nodes are active and only one type of traffic is allowed in the
network. In second category of experiment in section 7.3.2, 1000 calls are generated
while allowing traffic in half of the network, first in upper part and then in lower part
of the network; here again only one type of traffic is allowed to validate the results. In
the third category of experiment in section 7.3.3, all nodes are active and every time
102
two equal parts of traffic is allowed in the network: first 50% BE and 50% Class A;
then 50% BE and 50% Class B; and finally 50% BE and 50% Class C. Convergence is
shown at the end in section 7.3.4.
7.2.1 Network Response when all nodes are active
In this set of experiments 1000 calls are generated while keeping all the 50 nodes
active. Five tests have been performed and results are presented as under.
(i) Class A service only
Here the average hop is found as 5.80. This is because of the fact that as this service
requires more resources and lesser nodes are capable of providing these much
resources hence average is less in this case.
Num
ber o
f Hop
s
Route discovery requests
Figure 7.2. Hop Count while Only Class A Traffic
103
Packet drops in this case are as 416. This is because bandwidth requirement of this
service is more as compared to other types of traffic which MANET environment is
hard to provide, hence in this case more packet are dropped.
0
200
400
600
800
1000
1200
Total Successful Dropped
No.
of P
acke
ts
Figure 7.3 Packet Ratio of Class A
(ii) Class B service only
In this case only traffic belonging to class B is allowed in the network. Here the
average hop found is 5.90; obviously this is better as compared to the observations
made in case where only Class A traffic was present.
104
Num
ber o
f Hop
s
Route discovery requests
Figure 7.4. Hop Count while Only Class B Traffic
Figure 7.4 shows that average hop is more than the previous case because here more
nodes can provide the required resources. It is evident from Figure 7.5 that the Packet
drops are 325 which are clearly less than Class A as this service requires less resources
as compared to class A and the network can cater the need of Class B with more ease
than as compared with the previous where more resources were in demand.
105
0
200
400
600
800
1000
1200
Total Successful Dropped
No.
of P
acke
ts
Figure 7.5. Packet Ratio of Class B
(iii) Class C service only
Resource requirement of Class traffic is less than all other traffic quality of service
classes. Figure 7.6 shows the average hop in case where only traffic of class C is
allowed in the network; here average hop is 6.26 which is better than Class B and class
A. The rational behind is that more hops are at the level of providing this lower level
bandwidth.
106
Num
ber o
f Hop
s
Route discovery requests
Figure 7.6. Hop Count while Only Class C Traffic
Figure 7.7 shows the packet drop ratio where only class C traffic is allowed in the
network. Here Packet drops are at much lower side as at 265. This clearly indicates
that lesser traffic is blocked or not served in this scenario. This better performance is
due to the fact that now lesser resources are required as compared to the previous
cases.
107
0
200
400
600
800
1000
1200
Total Successful Dropped
No.
of P
acke
ts
Figure 7.7. Packet Ratio of Class C
(iv) Best Effort
In this case of only best effort type traffic present in the network. Two type of tests
have been performed to show the effectiveness of the proposed solution and
interesting observations are made.
Without QoS
In this case, no quality of service is applied in the network, this means that no
classification between traffic is done and all the traffic is allowed to be served by the
network without blocking any packet due to scarcity of resources. Figure 7.8 shows
the topology of this type of network. Here white circles show only best effort traffic
capable nodes, where at the moment no QoS is applied in the network.
108
Figure 7.8. Topology with no QoS
Num
ber o
f Hop
s
Route discovery requests
Figure 7.9. Hop Count while Only Best Effort Traffic with no QoS
109
Figure 7.9 indicates the average hop in this experiment which is found as 7.03. Figure
7.10 shows the Packet drops in case where only best effort traffic class is present in
the network without enabling quality of service in the network. Here packet drops are
43. Obviously the packet drops are less than all the previous experiment. This is
because resource requirement is extremely less in this case and network is ready to
serve almost all the traffic.
0
200
400
600
800
1000
1200
Total Successful Dropped
No.
of P
acke
ts
Figure 7.10. Packet Ratio of BE with no QoS
With QoS
This is an interesting case where only best effort class is served by the network but
enabling quality of service. This means that network is supposed to behave in the
fashion where classification of packets is carried though one type of class is allowed.
In other words here traffic is prioritized though no other classes are allowed at that
time. The traffic is of course best effort but prioritized instead. Figure 7.11 shows the
110
topology of this type of network; here nodes are shown green to denote QoS enabled
instead of white nodes as in the previous case.
Figure 7.11. Topology with QoS
It is interesting to note here is that now both these parameters are less than the
previous best effort case; this is because now it requires guaranteed services instead of
unreliable best effort.
111
Num
ber o
f Hop
s
Route discovery requests
Figure 7.12. Hop Count, Only Best Effort Traffic with QoS active
0
200
400
600
800
1000
1200
Total Successful Dropped
No.
of P
acke
ts
Figure 7.13. Packet Ratio of BE with QoS
Figure 7.12 shows the average hop slightly increased to 6.90 and packet drops is
slightly increased to 74.
112
7.2.2 Network response when half of topology active
In this set of experiments, 1000 calls are made while keeping only half of the topology
active at a time; this means that not all the nodes are active at a time rather half of the
nodes are relaying the traffic. Firstly upper part of the network is kept active and later
the lower part. Again only one type of traffic is allowed at a time for better
understandings of results. This category of tests is important because here upper part
of topology is having those 25 nodes which have no breakage of route, while lower 25
nodes are linked in such a manner that some nodes go out of the network coverage
giving rise to much higher packet drops.
(i) Test 1: Class A, Upper half
Here half of all the nodes carrying traffic of class A and rest of the nodes are inactive.
Figure 7.14 shows the topology made for this experiment. Here pink nodes are
showing class A traffic enabled nodes and rest blue nodes are inactive nodes.
113
Figure 7.14. Upper half nodes active for class A
Num
ber o
f Hop
s
Route discovery requests
Figure 7.15. Hop count upper half class A
114
In this experiment the average hop is found to be 5.19 and packet drops is found as
474. This is because nodes in upper part are well connected.
(ii) Test 2: Class A, Lower half
This experiment is reverse of the above test. Here the upper half part nodes are
inactive while lower 25 nodes are carrying the traffic of class A as shown in Figure
7.16, and the results are shown in Figure 7.17. It is found that the average hop is 2.07
and packet drops is 895. This behavior validates our algorithm as here nodes are
sparsely populated and are much apart and even out of network coverage as compared
to the upper half scenario which gives lesser packets drops.
Figure 7.16. Lower half nodes active for class A
115
Num
ber o
f Hop
s
Route discovery requests
Figure 7.17. Hop count Lower half class A
(iii) Test 3: Class B, Upper half
Two similar tests are performed for Class B traffic. Here half of all the nodes are
carrying traffic of class B only and rest of the nodes are inactive. Figure 7.18 shows
the resultant topology as Yellow nodes are showing class B capable nodes and Blue
are inactive nodes.
116
Figure 7.18. Upper half nodes active for class B
In this test the average hop is observed to be 4.99 and packet drop is found at the level
of 415 as in Figure 7.19. This is because upper half contains well connected nodes.
117
Num
ber o
f Hop
s
Route discovery requests
Figure 7.19. Hop count upper half class B
(iv) Test 4: Class B, Lower half
In this case lower nodes are made active and upper are made inactive as in Figure
7.20. The average hop count in Figure 7.21 is found at 2.22 and packet drop is found
as high as 873. This trend is exactly like previous class A, where upper half performed
much better than lower half one as lower half consists on not well connected nodes.
118
Figure 7.20. Lower Half topology active for class B
Num
ber o
f Hop
s
Route discovery requests
Figure 7.21. Hop count lower half class B
119
(v) Test 5: Class C, Upper half
On the pattern of previous two, the same test is performed for Class C. Here half of all
the nodes are carrying traffic of class C, these are in upper part of topology shown.
This is shown in Figure 7.22 as Grey nodes are showing class C capable nodes and
Blue are inactive nodes.
Figure 7.22. Upper Half topology active for class C
120
Num
ber o
f Hop
s
Route discovery requests
Figure 7.23. Hop count upper half class C
In this case again as in Figure 7.23 the average hop is found as 5.06 and packet drop is
found as 403. These results are as expected proving the effectiveness of this proposed
algorithm..
(vi) Test 6: Class C, Lower half
This test is the reverse of previous as now lower nodes of the topology are made active
and upper nodes are made inactive (Figure 7.24) and results are obtained in Figure
7.25.
121
Figure 7.24. Lower half topology active for class C
Num
ber o
f Hop
s
Route discovery requests
Figure 7.25. Hop count lower half class C
122
In this experiment, the results show average hop as 2.31 and packet drop is in the
range of 879; clearly this is much lower than its counterpart of upper topology and is
exactly on the pattern of previous classes tests of this category.
(vii) Test 7: Best Effort, Upper half
Here upper half of all the nodes carry traffic of Best Effort only and rest of the nodes
are inactive. In Figure 7.26 white nodes are showing Best effort capable nodes and
Blue are inactive nodes.
Figure 7.26. Upper Half topology active for Best Effort
123
Num
ber o
f Hop
s
Route discovery requests
Figure 7.27. Hop count upper half BE
Here in Figure 7.27 the Average hop is 5.86 and packet drop is 140 only, it shows that
this behavior will degrade drastically in the next test.
(viii) Test 8: Best Effort, Lower half
In this case, lower half have 25 nodes capable of carrying the traffic of best effort type
while upper 25 nodes are made inactive. This is clearly shown in Figure 7.28.
124
Figure 7.28. Lower Half topology active for BE
Num
ber o
f Hop
s
Route discovery requests
Figure 7.29. Hop count lower half BE
125
In this test, the average hop in Figure 7.29 is observed as 2.89 and the packet drop is in
the range of 854. Of course this is much worst than previous best effort case where
upper topology was active instead.
In the above category of tests, upper half and lower half topology were tested against
each class of traffic and it was found that upper half case outperformed than lower half
because in lower half the nodes are sparsely populated and sometimes even out of
coverage hence much packets are dropped.
7.2.3 Network Response with 50 percent QoS traffic and 50 percent BE
In this set of experiments, 1000 calls are made while allowing two type of traffic at the
same time as a combination of BE with Class A, Class B and Class C in exact
proportion. First experiment is performed with BE and Class A followed by BE with
Class B and finally BE with Class C in these test. Interesting observations have been
noted here for each class. In topology green nodes represent BE while pink, yellow
and grey denote Class A, Class B and Class C nodes respectively.
126
(i) Test 1: Class A + Class BE
Figure 7.30. Class A and BE nodes
Figure 7.30 shows the topology using combination of Class A and BE nodes. From
Figure 7.31, average hop is 6.74 and the total packet drop is 244, in which 48 belongs
to BE and 196 are of Class A. Obviously more packet drops belong to resource
demanding traffic.
127
Num
ber o
f Hop
s
Route discovery requests
Figure 7.31. Hop count of Class A and BE
(ii) Test 2: Class B + Class BE
In this test, two types of traffic is allowed in the network in exact proportion as best
effort and class B. The resultant topology is shown in Figure 7.32, in which yellow
nodes represent class B enabled nodes while green nodes depict best effort relaying
nodes.
128
Figure 7.32. Class B and BE nodes
Num
ber o
f Hop
s
Route discovery requests
Figure 7.33. Hop count of class B and BE
129
The results are shown in Figure 7.33 which gives average hop of 6.59 and packet
drops are 175, where 28 belong to BE and 147 are from Class B. It is interesting to
note that the ratio between BE and Class B is less as compared with previous Class A
test. This is due to the fact that class B requires lesser resources hence tends to be
nearer to BE.
(iii) Test 3: Class C + Class BE
Figure 7.34. Class C and BE nodes
In this test, 50 percent traffic belongs to Class C and 50 percent belongs to Best Effort
class. This is shown in Figure 7.34 where purple nodes depict class C enabled nodes
while green nodes represent the Best Effort class enabled nodes.
130
Num
ber o
f Hop
s
Route discovery requests
Figure 7.35. Hop count of Class C and BE
Here average hop is 6.51 and packet drop is 151, where 45 belongs to BE and 106
belong to Class C.
One interesting observation in this last category of experiments is that the ratio
between packet drops of a class and its corresponding BE is more in case of Class A
case while lowest in case of class C. This is because class C is nearer to BE as
compared to Class A in resource demanding.
7.2.4 Convergence
It is worth noting that once the network is converged, then even increasing quality of
service traffic will not decrease the network throughput because of the scalable nature
of swarm intelligence protocols. The last simulation clearly indicates that after
131
convergence, the QoS based traffic may enjoy a better average. Detailed graph
explaining each class separately with a comparison to old one is shown in Figure 8.36.
Num
ber o
f Hop
s
Route discovery requests
Figure 7.36. QoS classes before Convergence and after Convergence
While looking into the details of quality of service classes, it could be found that this
time number of failed requests decreased to a greater extent. This is because the
network has utilized the knowledge from previous simulations and has chosen the
paths which are more stable according to pheromone table developed so far. This is
presented in more elaborated form in Figure 7.37.
132
Num
ber
of H
ops
Successful Packets
Figure 7.37. Successful Packets before and after the convergence
7.3 Conclusion In this chapter simulation results have been presented for the proposed Ants Inspired
Traffic Classification in Mobile Ad Hoc Network. The topology used, consist of 50
randomly placed nodes. The paths joining these nodes are having different link
characteristic, this is to be as near as possible to real life scenarios. Link bandwidth is
as a metric to determine the QoS availability on these links.
133
CHAPTER 8
VALIDATION OF RESULTS
8.1 Introduction ANOVA tests are performed to validate the results obtained in the previous chapter.
30 tests have been performed for each class involving 30 simulations. Minitab is used
for running ANOVA tests and the graphs are plotted accordingly. In this process, we
used four classes including; Class A, Class B, Class C, and BE as Class D. These tests
consist of two parts namely (a) Validation of Data, (b) Analysis of Data
8.1.2. Validation of Data
In this part, the data obtained from the results through 30 simulations have been
validated. Four Tests were performed including:
Normality Test
Randomness Test
Histogram
Diagnostic Report of Individual Classes
The Normal Probability Plot in Figure 8.1 shows the normality of the data.
134
50250-25-50
99.9
99
9590
80706050403020
10
5
1
0.1
Residual
Perc
ent
Normal Probability Plot(responses are classA, class B, Class C, Class D)
Figure 8.1. Normal Probability Graph
135
Figure 8.2 shows the Randomness of data, the residuals are scattered randomly across zero.
5004003002001000
50
25
0
-25
-50
Fitted Value
Res
idua
lVersus Fits
(responses are classA, class B, Class C, Class D)
Figure 8.2. Randomness graph
136
Figure 8.3 shows the histogram. It is approximately bell shaped and symmetric as
well. This means that no outliers.
4530150-15-30-45
25
20
15
10
5
0
Residual
Freq
uenc
y
Histogram(responses are classA, class B, Class C, Class D)
Figure 8.3. Histogram
137
Diagnostic Report in Figure 8.4 shows the support for our hypothesis which shows
that mean packet drops of all the classes are not equal.
400
200
0
400
200
0
400
200
0
400
200
0
classA
class B
Class C
Class D
560480400320240160800
classA
class B
Class C
Class D
Data in Worksheet OrderInvestigate outliers (marked in red).
Distribution of DataCompare the location and spread.
One-Way ANOVA for classA, class B, Class C, Class DDiagnostic Report
Figure 8.4. Diagnostics of individual classes
138
Figure 8.5 shows the summary report of one way ANOVA tests, which shows the
difference among means of classes is significant.
Differences among the means are significant (p < 0.05).
> 0.50.10.050
NoYes
P = 0.000
classA
class B
Class C
Class D
4803602401200
implications.the size of the differences to determine if they have practicaloverlap indicate means that differ from each other. ConsiderChart to identify means that differ. Red intervals that do notmeans at the 0.05 level of significance. Use the ComparisonYou can conclude that there are differences among the
1 Class D 2 3 42 Class C 1 3 43 class B 1 2 44 classA 1 2 3
# Sample Differs fromWhich means differ?
One-Way ANOVA for classA, class B, Class C, Class DSummary Report
Do the means differ?
Means Comparison ChartRed intervals that do not overlap differ.
Comments
Figure 8.5. Summary Report
8.1.2. Analysis of Data
In the previous section validation of data was performed. This was done by validating
the hypothesis that the mean packet drops of all the classes are not equal. to show that
the mean packet drops of all the classes is in the order as it should be, it is meant that
139
the mean packet drop of Class A is higher than of Class B and mean packet drop of
Class B is higher than of Class C and that of Class C is higher than of Class D which
is Best Effort Class. For this Paired T-Tests are performed in a way that two classes
are compared at a time.
Test 1: Paired T-Test; Class A and Class B
class B (p < 0.05).The mean of classA is significantly greater than the mean of
> 0.50.10.050
NoYes
P = 0.000
20016012080400
0
interpreting the results of the test.differences to zero. Look for unusual differences before-- Distribution of differences: Compare the location of thethat the true difference is between 113.34 and 131.99.the difference from sample data. You can be 90% confident-- CI: Quantifies the uncertainty associated with estimatingthe paired differences is greater than zero.than class B at the 0.05 level of significance. The mean of-- Test: You can conclude that the mean of classA is greater
Sample size 30Mean 122.67 90% CI (113.34, 131.99)Standard deviation 30.057
Statistics Differences *Paired
Mean 506.27 383.6Standard deviation 22.249 15.046
classA class B
* The difference is defined as classA - class B.
Paired t Test for the Mean of classA and class BSummary Report
Mean TestIs classA greater than class B?
Distribution of the DifferencesWhere are the differences relative to zero?
Comments
Figure 8.6. Paired T-test between Class A and Class B
140
Test 2: Paired T-Test; Class B and Class C
Class C (p < 0.05).The mean of class B is significantly greater than the mean of
> 0.50.10.050
NoYes
P = 0.000
100806040200
0
interpreting the results of the test.differences to zero. Look for unusual differences before-- Distribution of differences: Compare the location of thethat the true difference is between 62.797 and 74.936.the difference from sample data. You can be 90% confident-- CI: Quantifies the uncertainty associated with estimatingmean of the paired differences is greater than zero.greater than Class C at the 0.05 level of significance. The-- Test: You can conclude that the mean of class B is
Sample size 30Mean 68.867 90% CI (62.797, 74.936)Standard deviation 19.566
Statistics Differences *Paired
Mean 383.6 314.73Standard deviation 15.046 16.088
class B Class C
* The difference is defined as class B - Class C.
Paired t Test for the Mean of class B and Class CSummary Report
Mean TestIs class B greater than Class C?
Distribution of the DifferencesWhere are the differences relative to zero?
Comments
Figure 8.7. Paired T-test between Class B and Class C
141
Test 3: Paired T-Test; Class C and Class D
of Class D (p < 0.05).The mean of Class C is significantly greater than the mean
> 0.50.10.050
NoYes
P = 0.000
28024020016012080400
0
interpreting the results of the test.differences to zero. Look for unusual differences before-- Distribution of differences: Compare the location of thethat the true difference is between 273.44 and 284.23.the difference from sample data. You can be 90% confident-- CI: Quantifies the uncertainty associated with estimatingmean of the paired differences is greater than zero.greater than Class D at the 0.05 level of significance. The-- Test: You can conclude that the mean of Class C is
Sample size 30Mean 278.83 90% CI (273.44, 284.23)Standard deviation 17.390
Statistics Differences *Paired
Mean 314.73 35.9Standard deviation 16.088 5.6223
Class C Class D
* The difference is defined as Class C - Class D.
Paired t Test for the Mean of Class C and Class DSummary Report
Mean TestIs Class C greater than Class D?
Distribution of the DifferencesWhere are the differences relative to zero?
Comments
Figure 8.8. Paired T-test between Class C and Class D
The above four T tests clearly establish our claim that the mean packet drops of Class
A is greater than mean packet drops of Class B, mean packet drops of Class B is
greater that mean packet drops of Class C, mean packet drops of Class C is greater that
mean packet drops of Class D which is best effort traffic.
142
CHAPTER 9
CONCLUSIONS AND FUTURE WORK
9.1 Conclusions In this thesis, routing in Mobile Ad Hoc networks has been thoroughly studied and
investigated by simulating leading routing protocols and ineffectiveness of
conventional protocols has been highlighted.
Swarm intelligence based routing algorithms are deriving researchers towards a new
unexplored area and are showing great potential. Therefore, a QoS aware routing
algorithm has been proposed which is based on swarm intelligence approach. The
proposed algorithm targets for Quality of Service provisioning in MANET and the
experimental results show the promise of the technique. Here packets are marked
according to four classes. Paths with higher pheromone concentration are chosen for
optimum route discovery through number of hops. The requests are dropped only in
cases where bandwidth requirement is not properly met. The results show the
effectiveness of Swarm intelligence technique towards the convergence of the solution
which enables us to offer better services than the best effort (BE) model and plays an
important role in providing QoS support in MANETs.
The ant agents’ contribution in complex stigmergic process in the proposed approach
gives rise to larger optimization solutions, presents self- regulatory routing patterns
and generates acceptable control overhead.
143
QoS paths must ensure path availability for real time traffic during dialogue sessions.
One of the techniques for this could be through already searched alternate routes.
Therefore the proposed algorithm incorporates this well and keep available alternate
routes during data sessions with good QOS metric.
Many of the proposed techniques are depended on lower layers feedbacks. The
proposed algorithm is independent Layer approach requiring no dependency on other
layers. Pheromone evaporation property is exploited well here in this proposed
solution to avoid stale routes.
9.2 Future Work There exists a lot of research work in the area of mobile ad networks. This reported
work is usually conducted on simulations rather than actual real life scenarios.
Although research work has given good solutions but real life testing of these
proposed protocols can be a potential future work.
Complexity and Congestion in the network arising through implementation of
proposed routing solutions in MANET can also be considered as an issue of future
research.
Search broadcast is a potential problem in resource limited Mobile Ad Hoc Networks
and consumes bandwidth in a larger way. Therefore avoiding this Scavenger can be a
good future research work.
144
REFERENCES
[1] Internet Engineering Task Force MANET Working group.
http://www.ietf.org/html.charters/manetcharter.html.
[2] MANET working grouphttp://datatracker.ietf.org/wg/manet.
[3] James Kennedy. Russell C. Eberhart, “Swarm Intelligence”, ISBN 1-55860-
595-9. 2001
[4] C. K. TOH, “Ad Hoc Mobile Wireless Networks”, Prentice Hall PTR, Upper
Saddle River, NJ, USA, 2002.
[5] C.E. Perkins, editor. “Ad Hoc Networking”, Addison-Wesley, 2001.
[6] Muhammad Hassan Nasir, “Gateway Discovery and handover solutions for
internet connectivity in wireless Ad hoc networks”, M.E thesis, Mehran UET
Jamshoro, Pakistan 2007.
[7] Dr. Mohamed Younis, Dr. Sebnem Zorlu Ozer, “Wireless Ad Hoc Networks:
Technologies and Challenges”, John wiley and Sons Ltd. Chichester, UK,
2006.
[8] Quality of service (QoS) concept and architecture (release 5) (TS23.107
v5.7.0). TS 23.107, 3GPP, 2002.
[9] Larry L. Peterson and Bruce S. Devi Computer Networks, “A System
Approach”, Morgan kaufmann 2003.
[10] Lina Narbutaite, Brunonas Dekeris, “IP QoS evaluation using interoperability
of differentiated and integrated services”, 4th WSEAS International
145
Conference on Electronics, Hardware, Wireless and Optical Communications
2005.
[11] S. chakrabarti and A. Mishra, “QoS issues in ad hoc wireless networks”,
Communications Magazine, IEEE, Vol. 39, No. 2, (2001), pp. 142-148. 2001
[12] C.Zhu, M.Corson, “QoS routing for mobile ad hoc networks”, IEEE Infocom
2002.
[13] L. M. Patnaik and Anand Suvernkar, “Multipath Routing to Provide Quality of
Service in Mobile Ad Hoc Networks”, Lecture Notes in Computer Science,
Volume 2571, pages 154-163, 2002.
[14] Michael Gerharz, Christian Vogt, Christian de Waal, “Current Approaches
Towards Improved Quality-of-Service Provision in Mobile Ad-hoc
Networks”, Technical Report, March 2003 Communications Systems,
Rheinische Friedrich Wilhelm University at Bonn, Germany.
[15] Yan He Abdel Wahab, “HQMM: A Hybrid QoS Model for Mobile Ad-hoc
Networks”, 11th IEEE Symposium on Computers and Communications ISCC,
2006.
[16] Chai-Keong Toh, “Ad Hoc Mobile Wireless Networks: Protocols and
Systems”, Prentice Hall, ISBN: 0-13-007817-4, 2002.
[17] A. S. Tanenbaum, “Computer Networks”, 4th Edition, Prentice Hall, Upper
Saddle River NJ, August 2002.
[18] P. Jacquet and L. Viennot, “Overhead in mobile ad-hoc network protocol”
Rapport de recherche 3965, INRIA, 2000.
146
[19] M. W. Chandra, “Extensions to OSPF to support mobile ad hoc networking”,
in Internet Engineering Task Force (IETF) draft, July 2004.
[20] Naserian, M. Tepe, K.E. Tarique, M., “Routing overhead analysis for
reactive routing protocols in wireless ad hoc networks”, IEEE WiMob 2005.
[21] S. Zhao, I. Seskar and D. Raychaudhuri, “Performance and scalability of self-
organizing hierarchical ad hoc wireless networks”, WCNC, wireless
Communications and Networking Conference, IEEE, pp 132-137, 2004.
[22] Mbarushimana, C.; Shahrabi, A, “Comparative Study of Reactive and
Proactive Routing Protocols Performance in Mobile Ad Hoc Networks”, IEEE
21st International Conference on Advanced Information Networking and
Applications Workshops, AINAW07, 2007.
[23] Babar Nazir, Azzam-ul-Asar, Rahmatullah Narejo, M . Masood Anwar
“Mobility and Load Analysis of DSDV and DSR Mobile Ad Hoc Routing
Protocols”, International Conference on Wireless Networks, USA, 2010, 12-
15 july 2010 Las Vegas, Nevada, USA
[24] Network Simulator-2
http://www.isi.edu/nsnam/ns.
[25] Charles E. Perkins, Elizabeth M. Belding-Royer, and Ian Chakeres. "Ad Hoc
On Demand Distance Vector (AODV) Routing." IETF Internet draft, draft-
perkins-manet-aodvbis-00.txt, Oct 2003.
147
[26] Papadimitratos, P.; Haas, Z.J. “Secure on-demand distance vector routing in
ad hoc networks”, IEEE/Sarnoff Symposium on Advances in Wired and
Wireless Communication, 2005.
[27] Krishna Gorantala, “Routing Protocols in Mobile Ad Hoc Networks” Thesis
in Computing Science, Umea University Sweeden. 2006.
[28] Namit Mishra, AmitPandey, Naveen Nahata, Rakesh Sinha, Dr. S. Tapaswi,
“Selection of Ad Hoc Network Routing Protocol for by Performance
Analysis”, COIT, 2008.
[29] http://www.rfc-editor.org/rfc/rfc4728.txt Feb. 2007.
[30] Jacquet, P. Muhlethaler, T. Clausen, A. Laouiti, A. Qayyum, L. Viennot,
“Optimized link state routing protocol for ad hoc networks”, IEEE INMIC
2001.
[31] Z. J. Haas, M. R. Pearlman, and P. Samar, "The zone routing protocol (ZRP)
for ad hoc networks," IETF, MANET Internet Draft, July 2002.
[32] P. Samar, M. R. perarlman and Z.J. Hass, “Independent zone routing: an
adaptive hybrid routing framework for ad hoc wireless networks”, IEEE
Transaction on Networking, vol. 12, pp. 595-608, 2004.
[33] Yihai Zhang Gulliver, T.A “Quality of service for ad hoc on-demand distance
vector routing”, IEEE Intl. Conf. on Wireless and Mobile Computing,
Networking and Communications (WiMob), 2005.
148
[34] I. Jawhar and J. Wu, “Quality of Sevice Routing in Mobile Ad Hoc
Networks,” Resource Management in Wireless Networking, M. Cardei, and
D. -Z. Du (eds.), Kluwer, 2004.
[35] CHEN Y-S and YU Y-T, “Spiral Multi-Path QoS Routing in a Wireless
Mobile Ad Hoc Network”, IEICE Transactions on Communications, E87-B
No.1, pp 104-116, 2004.
[36] Y. Ge, T. Kunz, L. Lamont, “Quality of Service Routing in Ad-Hoc Networks
Using OLSR,” HICSS 2003 presentation, Hawaii, USA, January 2003.
[37] E. Bonabeau, M. Dorigo and G. Theraulaz, “Swarm Intelligence: From
Natural to Artificial Systems”, Artificial Life, vol. 7, pp. 315-319, 2001.
[38] Deneubourg, J.-L., Theraulaz, G., and Beckers, R., "Swarm-made
architectures," European Conference on Artificial Life, pp. 123-133, 1992.
[39] N. R. Franks, “Army Ants: A Collective Intelligence”, American Scientist,
vol. 77, pp. 139-145, March-April 1989.
[40] E. Bonabeau, M. Dorigo and G. Theraulaz, “ Inspiration for Optimization
from social insect behaviour”, Nature, vol. 406, pp. 39-42, July 2000.
[41] Marco Dorigo, Thomas Stutzle, “Ant Colony Optimization”, MIT Press,
Prentice Hall, 2004.
[42] Paulo Eduardo Merloti , “Optimization Algorithms Inspired by Biological
Ants and Swarm Behavior” Cabo Bahia, Chula Vista, CA 91914 United States
of America, June 2004.
149
[43] Bonabeau E. and Meyer C., “Swarm Intelligence: A Whole New Way To
Think About Business”, Harvard Business Review; 79(5): 106-114, 2001.
[44] B. Baran and R. Sosa. “Antnet: Routing Algorithm for Data Networks based
on mobile Agents”. Argentine Symposium on Artificial Intelligence ASAI-
2000, Tandil – Argentina.
[45] Oida. K., Sekido, M. “ARS: an efficient agent-based routing system for QoS
guarantees”. Computer Communications (Elsevier) Vol. 23 (2000) 1437-1447.
[46] E. Gabber, M. A. Smith, “Trail Blazer: A Routing Algorithm Inspired by
Ants”, In Proc. 12th IEEE Int. Conf. on Network Protocols (ICNP), Germany,
2004.
[47] Marco Dorigo, Mauro Birattari, Christian Blum, Luca M. Gamberdella,
Francesco Mondada and Thomas Stutzle, “Ant Colony Optimization and
Swarm Intelligence:”, 4th International Workshop, ANTS 2004, Brussels,
Belgium, 2004.
[48] Bianchi, L. M. Gambardella, and M. Dorigo. Solving the homogeneous
probabilistic traveling salesman problem by the ACO metaheuristic. In M.
Dorigo, G. Di Caro, and M. Sampels, editors,Ants Algorithms - Proceedings
of ANTS 2002, Third International Workshop on Ant Algorithms, volume
2463 of Lecture Notes in Computer Science, pages 176ñ187. Springer-Verlag,
2002.
150
[49] E. L. Lawler and A. H. Rinnooy-Kan. ”The Traveling Salesman Problem: A
Guided Tour of Combinatorial Optimization”, Wiley-Interscience Series in
Discrete Mathematics, 2000.
[50] H. Yun and A. N. Zincir-Heywood, “Intelligent Ants for Adaptive Network
Routing”, Second Annual Conference on Communication Networks and
Services Research, pp 255-261, 2004.
[51] H. Matsuo and K. Mori, “Accelerated Ants Routing in Dynamic Networks”,
in proc. Intl. Conf. on Software Engineering, Artificial Intelligence,
Networking and Parallel/Distributed Computing, pp. 333-339, Aug 2001.
[52] Parunak H and Brueckner S., “Stigmergic Learning for Self-Organizing
Mobile Ad-Hoc Networks (MANETS)”, Third Int. Joint Conf. on
Autonomous Agents and Multi-Agent Systems, pp. 1324-1325, 2004.
[53] Yanxiang He, Weidong Wen, Hui Jin and Haowen Liu, “Ant Based Core-
Extraction Routing Protocol on Ad Hoc Wireless Networks”, IEEE 2005.
[54] G. Di Caro. A society of ant-like agents for adaptive routing in networks.
DEA thesis in Applied Sciences, Polytechnic School, Universit´e Libre de
Bruxelles, Brussels (Belgium), May 2001.
[55] G. Di Caro, F. Ducatelle, and L.M. Gambardella, “AntHocNet: an ant-based
hybrid routing algorithm for mobile ad hoc networks”, In Proceedings of
Parallel Problem Solving from Nature (PPSN) VIII, volume 3242 of Lecture
Notes in Computer Science, pages 461.470. Springer-Verlag, 2004.
151
[56] G. Di Caro, F. Ducatelle, and L.M. Gambardella, “AntHocNet: probabilistic
multi-path routing in mobile ad hoc networks using ant-like agents”,
Technical Report 16-04, IDSIA, Lugano (Switzerland), April 2004.
[57] G. Di Caro, F. Ducatelle, and L.M. Gambardella, “AntHocNet: an adaptive
nature-inspired algorithm for routing in mobile ad hoc networks”, European
Transactions on Telecommunications, 2005.
[58] Nazir, B., Ahmed, J., Khan, A., Azzam-ul-Asar, “Stigmergy of Ants for QoS
Provisioning in Dynamic Ad Hoc Networks”, International Conference on
Emerging Technologies, 2006. ICET06.
[59] Babar Nazir, Jamal Ahmed, Azzam-ul-Asar, “Ants Inspired Traffic
Classification in Mobile Ad Hoc Networks’, Mehran University Research
Journal of Engineering and Technology, 2008.
[60] M. Dorigo and G. Di Caro, “The ant colony optimization meta-heuristic”,
New ideas in Optimization, F. Glover, Ed. London, UK: McGraw-Hill, 1999.
152
Appendix DSR Otcl Code
#===================================
# Simulation parameters setup example
#===================================
set val(channel) WirelessChannel
set val(propagation) TwoRayGround
set val(net-if) WirelessPhy
set val(maclayer) 802.11
set val(ifque) DropTail
set val(ll) LL
set val(anttena) OmniAntenna
set val(ifqlength) 40
set val(nnodes) 40
set val(rtpl) DSR
set val(a) 1020
set val(b) 541
set val(stop) 20
#===================================
# Initialization
#===================================
#Creation of simulation
set ns [new Simulator]
#Parameters for setting up
set topo [new]
$topo load_flatgrid $val(a) $val(b)
153
create-god $val(nn)
# For tracing
set tracefile [openwireless DSR2.tr]
$ns trace-all $tracefile
#For visualization
set namfile [open wirelessDSR2.nam]
$ns namtrace-all $namfile
$ns namtrace-all-wireless $namfile $val(x) $val(y)
set chan [new $val(channel)]
# For channel creation
#----------------------------------------------------------------------
# Node related
#-----------------------------------------------------------------------
$ns node-config -adhocRouting $val(rtpl) \
-llType $val(ll) \
-macType $val(maclayer) \
-ifqType $val(ifqque) \
-ifqLen $val(ifqlength) \
-antType $val(anttena) \
-propType $val(propagation) \
-phyType $val(net-if) \
-channel $chan \
-topoInstance $topo \
-agentTrace ON \
-routerTrace ON \
-macTrace ON \
154
-movementTrace ON
#----------------------------------------------------------------------
# Defining node
#----------------------------------------------------------------------
#50 Nodes creation
set m0 [$ns node]
$m0 set A_ 600
$m0 set B_ 307
$m0 set C_ 0
$ns initial_node_pos $m0 30
set m [$ns node]
$m1 set A_ 724
$m1 set B_ 55
$m1 set C_ 0
$ns initial_node_pos $m1 30
set m2 [$ns node]
$m2 set A_ 535
$m2 set B_ 211
$m2 set C_ 0
$ns initial_node_pos $m2 30
set m3 [$ns node]
$m3 set A_ 545
$m3 set B_ 241
$m3 set C_ 0
$ns initial_node_pos $m3 30
set m4 [$ns node]
$m4 set A_ 490
$m4 set B_ 110
155
$m4 set C_ 0
$ns initial_node_pos $m4 30
set m5 [$ns node]
$m5 set A_ 761
$m5 set B_ 239
$m5 set C_ 0
$ns initial_node_pos $m5 30
set m6 [$ns node]
$m6 set A_ 829
$m6 set B_ 385
$m6 set C_ 0
$ns initial_node_pos $m6 30
set m7 [$ns node]
$m7 set A_ 820
$m7 set B_ 294
$m7 set C_ 0
$ns initial_node_pos $m7 30
set m8 [$ns node]
$m8 set A_ 446
$m8 set B_ 348
$m8 set C_ 0
$ns initial_node_pos $m8 30
set m9 [$ns node]
$m9 set A_ 518
$m9 set B_ 211
$m9 set C_ 0
$ns initial_node_pos $m9 30
set m10 [$ns node]
$m10 set A_ 723
156
$m10 set B_ 361
$m10 set C_ 0
$ns initial_node_pos $m10 30
set m11 [$ns node]
$m11 set A_ 858
$m11 set B_ 302
$m11 set C_ 0
$ns initial_node_pos $m11 30
set m12 [$ns node]
$m12 set A_ 450
$m12 set B_ 65
$m12 set C_ 0
$ns initial_node_pos $m12 30
set m13 [$ns node]
$m13 set A_ 618
$m13 set B_ 273
$m13 set C_ 0
$ns initial_node_pos $m13 30
set m14 [$ns node]
$m14 set A_ 465
$m14 set B_ 155
$m14 set C_ 0
$ns initial_node_pos $m14 30
set m15 [$ns node]
$m15 set A_ 628
$m15 set B_ 378
$m15 set C_ 0
$ns initial_node_pos $m15 30
set m16 [$ns node]
157
$m16 set A_ 729
$m16 set B_ 297
$m16 set C_ 0
$ns initial_node_pos $m16 30
set m17 [$ns node]
$m17 set A_ 742
$m17 set B_ 9
$m17 set C_ 0
$ns initial_node_pos $m17 30
set m18 [$ns node]
$m18 set A_ 690
$m18 set B_ 184
$m18 set C_ 0
$ns initial_node_pos $m18 30
set m19 [$ns node]
$m19 set A_ 861
$m19 set B_ 222
$m19 set C_ 0
$ns initial_node_pos $m19 30
set m20 [$ns node]
$m20 set A_ 529
$m20 set B_ 342
$m20 set C_ 0
$ns initial_node_pos $m20 30
set m21 [$ns node]
$m21 set A_ 642
$m21 set B_ 4
$m21 set C_ 0
$ns initial_node_pos $m21 30
158
set m22 [$ns node]
$m22 set A_ 792
$m22 set B_ 185
$m22 set C_ 0
$ns initial_node_pos $m22 30
set m23 [$ns node]
$m23 set A_ 675
$m23 set B_ 373
$m23 set C_ 0
$ns initial_node_pos $m23 30
set m24 [$ns node]
$m24 set A_ 450
$m24 set B_ 380
$m24 set C_ 0
$ns initial_node_pos $m24 30
set m25 [$ns node]
$m25 set A_ 452
$m25 set B_ 189
$m25 set C_ 0
$ns initial_node_pos $m25 30
set m26 [$ns node]
$m26 set A_ 602
$m26 set B_ -47
$m26 set C_ 0
$ns initial_node_pos $m26 30
set m27 [$ns node]
$m27 set A_ 906
$m27 set B_ 303
$m27 set C_ 0
159
$ns initial_node_pos $m27 30
set m28 [$ns node]
$m28 set A_ 865
$m28 set B_ 101
$m28 set C_ 0
$ns initial_node_pos $m28 30
set m29 [$ns node]
$m29 set A_ 611
$m29 set B_ -66
$m29 set C_ 0
$ns initial_node_pos $m29 30
set m30 [$ns node]
$m30 set A_ 456
$m30 set B_ -5
$m30 set C_ 0
$ns initial_node_pos $m30 30
set m31 [$ns node]
$m31 set A_ 471
$m31 set B_ 381
$m31 set C_ 0
$ns initial_node_pos $m31 30
set m32 [$ns node]
$m32 set A_ 559
$m32 set B_ 381
$m32 set C_ 0
$ns initial_node_pos $m32 30
set m33 [$ns node]
$m33 set A_ 807
$m33 set B_ 386
160
$m33 set C_ 0
$ns initial_node_pos $m33 30
set m34 [$ns node]
$m34 set A_ 847
$m34 set B_ 70
$m34 set C_ 0
$ns initial_node_pos $m34 30
set m35 [$ns node]
$m35 set A_ 786
$m35 set B_ 181
$m35 set C_ 0
$ns initial_node_pos $m35 30
set m36 [$ns node]
$m36 set A_ 858
$m36 set B_ 393
$m36 set C_ 0
$ns initial_node_pos $m36 30
set m37 [$ns node]
$m37 set A_ 503
$m37 set B_ 427
$m37 set C_ 0
$ns initial_node_pos $m37 30
set m38 [$ns node]
$m38 set A_ 508
$m38 set B_ -49
$m38 set C_ 0
$ns initial_node_pos $m38 30
set m39 [$ns node]
$m39 set A_ 815
161
$m39 set B_ 405
$m39 set C_ 0
$ns initial_node_pos $m39 30
set m40 [$ns node]
$m40 set A_ 641
$m40 set B_ -51
$m40 set C_ 0
$ns initial_node_pos $m40 30
set m41 [$ns node]
$m41 set A_ 708
$m41 set B_ 1
$m41 set C_ 0
$ns initial_node_pos $m41 30
set m42 [$ns node]
$m42 set A_ 856
$m42 set B_ 127
$m42 set C_ 0
$ns initial_node_pos $m42 30
set m43 [$ns node]
$m43 set A_ 516
$m43 set B_ 138
$m43 set C_ 0
$ns initial_node_pos $m43 30
set m44 [$ns node]
$m44 set A_ 756
$m44 set B_ 214
$m44 set C_ 0
$ns initial_node_pos $m44 30
set m45 [$ns node]
162
$m45 set A_ 804
$m45 set B_ 431
$m45 set C_ 0
$ns initial_node_pos $m45 30
set m46 [$ns node]
$m46 set A_ 910
$m46 set B_ 41
$m46 set C_ 0
$ns initial_node_pos $m46 30
set m47 [$ns node]
$m47 set A_ 525
$m47 set B_ -23
$m47 set C_ 0
$ns initial_node_pos $m47 30
set m48 [$ns node]
$m48 set A_ 474
$m48 set B_ 329
$m48 set C_ 0
$ns initial_node_pos $m48 30
set m49 [$ns node]
$m49 set A_ 841
$m49 set B_ 220
$m49 set C_ 0
$ns initial_node_pos $49 30
#----------------------------------------------------------------------
# Finishing
#-----------------------------------------------------------------------
#Finishing process
163
proc finish {} {
global ns tracefile namfile
$ns flush-trace
close $tracefile
close $namfile
exec nam wirelessDSR2.nam &
exit 0
}
for {set i 0} {$i < $val(nn) } { incr i } {
$ns at $val(stop) "\$n$i reset"
}
$ns at $val(stop) "$ns nam-end-wireless $val(stop)"
$ns at $val(stop) "finish"
$ns at $val(stop) "puts \"done\" ; $ns halt"
$ns run
164
DSDV Otcl Code
#By using user defind patterns
# -------------------------------------------------------------
# Initial Setup
# -------------------------------------------------------------
set val(channel) WirelessChannel
set val(propagation) TwoRayGround
set val(net-if) WirelessPhy
set val(maclayer) 802.11
# Simulation setup
set val(ifq) DropTail
set val(ll) LL
set val(antenna) OmnidirAntenna
set val(a) 1000
set val(b) 1000
set val(ifqlength) 60
set val(seed) 1.0
set val(adhocRouting) DSDV
set val(nnodes) 50
set val(cp) "/root/ns-allinone-2.29/ns-2.29/indep-utils/cmu-scen-gen/cbr-50-2"
set val(sc) "/root/ns-allinone-2.29/ns-2.29/indep-utils/cmu-scen-gen/setdest/scen-50-
5"
set val(stop) 200
# Core code here
165
# Initialization
set ns_ [new Simulator]
# Defining topology
set topo [new]
set tracefd [open wireless2-DSDV.tr]
set namtrace [open wireless2-DSDV.nam]
$ns_ trace-all $tracefd
$ns_ namtrace-all-wireless $namtrace $val(a) $val(b)
$topo load_flatgrid $val(a) $val(b)
set god_ [create-god $val(nn)]
# defining nodes
set chan [new $val(channel)]
$ns_ node-config -adhocRouting $val(adhocRouting) \
-llType $val(ll) \
-macType $val(maclayer) \
-ifqType $val(ifq) \
-ifqLen $val(ifqlength) \
-antType $val(antenna) \
-propType $val(propagation) \
-phyType $val(net-if) \
-channel $chan \
-topoInstance $topo \
-agentTrace ON \
-routerTrace ON \
166
-macTrace ON
# creation of nodes
for {set i 0} {$i < $val(nn) } {incr i} {
set node_($i) [$ns_ node]
$node_($i) random-motion 0
}
puts "coming up the connectivity model"
source $val(cp)
puts "coming up the scenario model"
source $val(sc)
for {set i 0} {$i < $val(nn)} {incr i} {
$ns_ initial_node_pos $node_($i) 20
}
for {set i 0} {$i < $val(nn) } {incr i} {
$ns_ at $val(stop).0 "$node_($i) reset";
}
$ns_ at $val(stop).0002 "puts \"NS EXITING...\" ; $ns_ halt"
puts $tracefd "M 0.0 nn $val(nn) x $val(a) y $val(b) rp $val(adhocRouting)"
puts $tracefd "M 0.0 sc $val(sc) cp $val(cp) seed $val(seed)"
puts $tracefd "M 0.0 prop $val(propagation) ant $val(antenna)"
puts "Start of Simulator."
$ns_ run
167
Matlab Code % Project work %****************LOAD ANALYSIS**************** clc; close all; clear all; load=512*8*[0.1;2;4;8;10]; l=size(load); % Data uploading......... dsdv_recv_load=textread('dsdv_recv_load.txt','%n'); dsr_recv_load=textread('dsr_recv_load.txt','%n'); dsdv_drop_load=textread('dsdv_drop_load.txt','%n'); dsr_drop_load=textread('dsr_drop_load.txt','%n'); dsr_pkt_load=textread('dsr_pkt_load.txt','%n'); dsdv_pkt_load=textread('dsdv_pkt_load.txt','%n'); % Computational work for i=1:l % DSDV Calculation of Drop Rate and Through put dsdv_load(i)=dsdv_recv_load(i)+dsdv_drop_load(i); DSDVdropRatio(i)=dsdv_drop_load(i)/dsdv_load(i); DSDVthroughput(i)=dsdv_recv_load(i)/dsdv_load(i); % DSR Calculation of Drop Rate and Through put dsr_load(i)=dsr_recv_load(i)+dsr_drop_load(i); DSRdropRatio(i)=dsr_drop_load(i)/dsr_load(i); DSRthroughput(i)=dsr_recv_load(i)/dsr_load(i); %Over head computation dsr_overhead(i)=32*dsr_pkt_load(i)/dsr_load(i); dsdv_overhead(i)=32*dsdv_pkt_load(i)/dsdv_load(i); end % Packet Drop Ratio Analysis DSR & DSDV Protocols plot(load,DSRdropRatio,'bo-'); hold on plot(load,DSDVdropRatio,'ro:'); grid on xlabel('Data Rate[kbps]'); ylabel('DropRatio'); title('Load Analysis'); legend('DSR','DSDV'); % Through put analysis of DSR & DSDV protocols figure plot(load,DSRthroughput,'b+-'); hold on plot(load,DSDVthroughput,'r+:'); grid on xlabel('Data Rate[Kbps]');
168
ylabel('Throughput'); title('Load Analysis'); legend('DSR','DSDV'); % Protocol Overhead Analysis figure plot(load,dsr_overhead,'bo-'); hold on plot(load,dsdv_overhead,'ro:'); grid on xlabel('Data Rate[kbps]'); ylabel('Overhead'); title('Protocol Overhead Analysis'); legend('DSR','DSDV'); % end % Project work %****************moblity ANALYSIS**************** clc; close all; clear all; moblity=[2;5;10;15;20]; m=size(moblity); % Data uploading......... dsdv_recv_moblity=textread('dsdv_recv_moblity.txt','%n'); dsr_recv_moblity=textread('dsr_recv_moblity.txt','%n'); dsdv_drop_moblity=textread('dsdv_drop_moblity.txt','%n'); dsr_drop_moblity=textread('dsr_drop_moblity.txt','%n'); dsr_pkt_moblity=textread('dsr_pkt_moblity.txt','%n'); dsdv_pkt_moblity=textread('dsdv_pkt_moblity.txt','%n'); % Computational work for i=1:m % DSDV Calculation of Drop Rate and Through put dsdv_moblity(i)=dsdv_recv_moblity(i)+dsdv_drop_moblity(i); DSDVdropRatio(i)=dsdv_drop_moblity(i)/dsdv_moblity(i); DSDVthroughput(i)=dsdv_recv_moblity(i)/dsdv_moblity(i); % DSR Calculation of Drop Rate and Through put dsr_moblity(i)=dsr_recv_moblity(i)+dsr_drop_moblity(i); DSRdropRatio(i)=dsr_drop_moblity(i)/dsr_moblity(i); DSRthroughput(i)=dsr_recv_moblity(i)/dsr_moblity(i); %Over head computation dsr_overhead(i)=32*dsr_pkt_moblity(i)/dsr_moblity(i); dsdv_overhead(i)=32*dsdv_pkt_moblity(i)/dsdv_moblity(i); end % Packet Drop Ratio Analysis DSR & DSDV Protocols plot(moblity,DSRdropRatio,'bo-');
169
hold on plot(moblity,DSDVdropRatio,'ro:'); grid on xlabel('Moblity Speed[m/sec]'); ylabel('DropRatio'); title('moblity Analysis'); legend('DSR','DSDV'); % Through put analysis of DSR & DSDV protocols figure plot(moblity,DSRthroughput,'b+-'); hold on plot(moblity,DSDVthroughput,'r+:'); grid on xlabel('Moblity Speed[m/sec]'); ylabel('Throughput'); title('moblity Analysis'); legend('DSR','DSDV'); % Protocol Overhead Analysis figure plot(moblity,dsr_overhead,'bo-'); hold on plot(moblity,dsdv_overhead,'ro:'); grid on xlabel('Moblity speed[m/sec]'); ylabel('Overhead'); title('Protocol Overhead Analysis'); legend('DSR','DSDV'); % end