modeling and analytical performance evaluation of route
TRANSCRIPT
i
Modeling and Analytical Performance Evaluation of Route Discovery/Route Maintenance in Reactive and
Proactive Routing Protocols for MANETs and VANETs
By
Mr. Danish Mahmood
Registration Number: CIIT/FA10-REE-010/ISB
MS Thesis
In
Network Engineering
Electrical Engineering Department
COMSATS Institute of Information Technology
Islamabad β Pakistan
Fall, 2011
ii
COMSATS Institute of Information Technology
Modeling and Analytical Performance Evaluation of Route Discovery/Route Maintenance in Reactive and
Proactive Routing Protocols for MANETs and VANETs
A Thesis Presented to
COMSATS Institute of Information Technology, Islamabad
In partial fulfillment
Of the requirement for the degree of
MS (Network Engineering)
By
Mr. Danish Mahmood
CIIT/FA10-REE-010/ISB
Fall, 2011
iii
Modeling and Analytical Performance Evaluation of Route discovery/Route Maintenance in Reactive and Proactive Routing Protocols for MANETs and VANETs
A post Graduate Thesis submitted to Electrical Engineering as partial fulfillment of the requirement for the award of Degree of M.S (Electrical Engineering).
Name Registration Number
Mr. Danish Mahmood CIIT/FA10-REE-010 /ISB
Supervisor
Dr. Nadeem Javaid
Assistant Professor
Department of Electrical Engineering
Islamabad Campus
COMSATS Institute of Information Technology (CIIT)
Islamabad Campus
December, 2011
iv
Final Approval
This thesis titled
Modeling and Analytical Performance Evaluation of Route Discovery/Route Maintenance in Reactive and
Proactive Routing Protocols for MANETs and VANETs By
Mr. Danish Mahmood
CIIT/FA10-REE-010/ISB
Has been approved
For the COMSATS Institute of Information Technology, Islamabad
External Examinar: __________________________________
Supervisor: ________________________
Dr.Nadeem Javaid Department of Electrical Engineering Islamabad Campus
HoD: ________________________
Dr. ShafayatAbrar Department of Electrical Engineering Islamabad Campus
v
Declaration
I Mr. Danish Mahmood, CIIT/FA10-REE-010/ISB hereby declare that I have produced the work presented in this thesis, during the scheduled period of study. I also declare that I have not taken any material from any source except referred to wherever due that amount of plagiarism is within acceptable range. If a violation of HEC rules on research has occurred in this thesis, I shall be liable to punishable action under the plagiarism rules of the HEC.
Date: ________________ Signature of the student:
________________ Mr. Danish Mahmood CIIT/FA10-REE-051/ISB
vi
Certificate
It is certified that Mr. Danish Mahmood CIIT/FA10-REE-051/ISB has carried out all the work related to this thesis under my supervision at the Department of Electrical Engineering COMSATS Institute of Information Technology, Islamabad and the work fulfills the requirements for award of MS degree.
Date: _________________
Supervisor: ____________________
Dr. Nadeem Javaid Department of Electrical Engineering CIIT Islamabad Campus
Head of Department:
_____________________________
Dr.Shafayat Abrar HoD Electrical Engineering
vii
DEDICATION My work is dedicated to:
To ALLAH Almighty for His guidance and mercy over me,
To my family for their immense love and moral support throughout my masterβs program,
To my babies Raaim Mahmood and Nofil Mahmood (Son and Nephew),
Thanks to Almighty that I am blessed with such parents who always backed me up
And
The person who made this work possible, my thesis supervisor, Dr. Nadeem Javaid who acted not only as a supervisor but on every step of my work, he is the one who stood besides and supported everywhere with the help of his knowledge, guidance and any kind of support I needed.
viii
ACKNOWLEDGMENT
I am grateful to my supervisor, Dr. Nadeem Javaid, whose gradual encouragement, guidance and insightful criticism from the beginning to the final level enabled me have a deep understanding of the thesis. My Thesis Supervisor acted not only as a supervisor but on every step of my work, he is the one who stood besides and supported everywhere with the help of his knowledge, guidance and every kind of support I needed.
Dr. Safdar H. Bouk and Dr. Mustafa Shakir must also be thanked as they, on steps have spared their valuable time for discussing thesis and motivation to work harder.
My Lab Team, must also be acknowledged as it is all due to their cooperation and support that I managed to work.
And
Almighty Allah, who has blessed me with such learned parents, learned supervisor and gave me opportunity to sit besides learned scholars. Without his blessings nothing was possible.
Mr. Danish Mahmood
CIIT/FA10-REE-051/ISB
ix
ABSTRACT
Ad Hoc networks are the burning issue amongst researchers in current era. The reason is its
challenging environment where a solid mechanism is needed to provide efficient and reliable
communications. Mobility of wireless station which we refer as nodes along with the wireless
communication issues in combination with several other limitations gives a challenging
environment for working on Ah Hoc networks. MANETs or VANETS both are Ad Hoc
networks. In MANETs there may be mobility but not that much as one can imagine in
VANETs. Hence every scenario requires some different attributes to make communication
between nodes possible. Routing protocols play a vital role as they are responsible of
directing information from one sending node to other receiving node. For static and barely
mobile environment, Proactive routing protocols have developed while for highly mobile
environment, reactive protocols are proposed. There are Hybrid protocols as well but, in this
Thesis we confine ourselves to Route discovery and maintenance phases of proactive and
reactive routing protocols which are widely studied and observed now days. We observed
that less mathematical modeling is done with respect to simulations. Hence we contributed a
generalized analytical model for routing overhead of route discovery and maintenance phases
of both proactive and reactive routing protocols. Besides analytical modeling of reactive and
proactive routing protocols, we choose most widely studied reactive routing protocols i.e.
AODV, DSR and DYMO and compare them with each other and most prominent proactive
routing protocols i.e. FSR, DSDV and OLSR. We simulate these six routing protocols under
metrics of throughput, End - End delay and routing over head and discuss results keeping
mobility and scalability factor in emphasis.
x
TABLE OF CONTENTS
1. INTRODUCTION β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦β¦..1 1.1 Taxonomy of Routing Protocolβ¦...β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦3 1.2 Proactive Routing Protocolsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦..β¦β¦β¦.4 1.3 Reactive Routing Protocolsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦..β¦β¦......4
1.3.1 Route Discovery Phaseβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...5 1.3.2 Route Maintenance Phaseβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...5 1.3.3 Route Deletionβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦6
2. Routing: Reactive & Proactiveβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..........β¦β¦β¦..7 2.1 Reactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦β¦β¦β¦8 2.2 Ad Hoc On Demand Distance Vector (AODVβ¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦..β¦β¦8
2.2.1 Route Discovery and Route Maintenance of AODVβ¦β¦β¦β¦β¦β¦β¦...10 2.2.2 Binary Exponential Back off Algorithm and AODVβ¦β¦β¦β¦β¦β¦β¦...13 2.2.3 BEB Algorithm for AODVβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...14 2.2.4 Expanding Ring Search Algorithm and AODVβ¦β¦β¦β¦β¦β¦β¦.β¦β¦..14 2.2.5 ERS Algorithm for AODVβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦..15
2.3 Dynamic Source Routing β¦β¦.....β¦β¦β¦β¦β¦β¦β¦β¦..β¦................................16 2.3.1 Route Discovery and Route Maintenance of DSRβ¦β¦β¦β¦β¦β¦β¦β¦...16 2.3.2 Binary Exponential Back Off Algorithm and DSRβ¦β¦β¦β¦β¦β¦β¦β¦..19 2.3.3 Expanding Ring Search Algorithm and DSRβ¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦.20
2.4 Dynamic MANET On Demand (DYMO)β¦β¦β¦β¦β¦β¦β¦...β¦β¦β¦β¦β¦β¦β¦..20 2.4.1 Route Discovery and Route Maintenance of DYMOβ¦β¦β¦β¦β¦β¦...β¦21 2.4.2 BEB and ERS in DYMOβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦β¦23 2.4.3 BEB Algorithm for DYMOβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦25
2.5 Comparison: ERS and BEB in AODV, DSR and DYMOβ¦β¦β¦β¦β¦β¦β¦β¦β¦.26 2.6 Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦.27 2.7 Fisheye State Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.27
2.7.1 Fisheye Techniqueβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦28 2.7.2 FSR workingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦β¦β¦30
2.8 Destination Sequenced Distance Vector (DSDV)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦31 2.8.1 Route Discovery and Route Maintenance for DSDVβ¦β¦β¦β¦β¦β¦...β¦..31
2.9 Optimized Link State Routing (OLSR)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.34 2.9.1 Multipoint Relay Conceptβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦35 2.9.2 Route Discovery and Route Maintenance of OLSRβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦36
3. Analytical Modeling: Reactive and Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦.38 3.1 Reactive Modelingβ¦β¦. β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦.β¦β¦...39
3.1.1 Route Discoveryβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦.β¦β¦39 3.1.2 Route Maintenanceβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦39 3.1.3 Steps involved in RD and RMβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.40 3.1.4 Routing Overhead for a single Routeβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..40
xi
3.1.5 Overhead due to RREQ Packetsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦.41 3.1.6 Overhead due to RREP Packetsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦β¦43 3.1.7 Monitoring Overhead of one Linkβ¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦β¦β¦β¦44 3.1.8 Routing Overhead for βnβ Routesβ¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦β¦β¦β¦β¦β¦.45
3.2 Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.47 3.2.1 Route failure Impact on routing overheadβ¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦.48 3.2.2 Periodic Message Overheadβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦49 3.2.3 Triggered Message Overheadβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦β¦.49 3.2.4 Routing Overhead of Proactive Routing Protocolβ¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦..51 3.2.5 Calculations and Resultsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦.53
4. Experiments, Results and Discussionsβ¦β¦β¦β¦β¦...β¦β¦β¦β¦β¦β¦β¦..β¦.β¦β¦β¦57 4.1 Simulationsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦58 4.2 Parameters for Experimentsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦58 4.3 Reactive Experimentsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦.59
4.3.1 Throughput of Reactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦β¦β¦..59 4.3.2 End - End Delay of Reactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦.61 4.3.3 Normalized Routing Load of Reactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦62
4.4 Proactive Experimentsβ¦β¦β¦β¦β¦...β¦......β¦.β¦β¦......................................β¦β¦β¦64 4.4.1 Throughput of Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦.64 4.4.2 End - End Delay of Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..65 4.4.3 Normalized Routing Load of Proactive Routingβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦67
4.5 Conclusionβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦...β¦β¦β¦β¦β¦β¦β¦......β¦β¦.68 5. Referencesβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦..β¦..β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦72
xii
LIST OF FIGURES
Figure 1.1 Routing protocolsβ¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦3
Figure 2.1 Propagation of RREQ packets in AODVβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦11
Figure 2.2 Propagation of RREP packets in AODVβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦β¦β¦β¦.12
Figure 2.3 Propagation of RREQ packets in DSRβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.17
Figure 2.4 Propagation of RREP packets in DSRβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.18
Figure 2.5 Propagation of RREQ packets in DYMOβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦21
Figure 2.6 Propagation of RREP packets in DYMOβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.22
Figure 2.7 Fisheye Routing Protocolβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.29
Figure 2.8 Triggered update notificationβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦β¦32
Figure 2.9 DSDV protocolβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...β¦β¦β¦..34
Figure 2.10 Functions of MPR in OLSRβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦β¦β¦.35
Figure 3.1 RREQ Flooding in Reactive Routing Protocolsβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦.42
Figure 3.2 RREP and link establishment in Reactive Routing Protocolsβ¦β¦β¦β¦β¦β¦β¦β¦..43
Figure 3.3 Link Monitoring in reactive Routing Protocol (AODV)β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦.45
Figure 3.4 Effect of triggered Messagesβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦..50
Figure 4.1 Throughput (AODV, DSR, DYMO)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦..61
Figure 4.2 End - End Delay (AODV, DSR, DYMO)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦63
Figure 4.3 Normalized Routing Load (AODV, DSR, DYMO)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦..64
Figure 4.4 Throughput (FSR. DSDV, OLSR)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..66
Figure 4.5 End - End Delay (FSR, DSDV, OLSR)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.67
Figure 4.6 Normalized Routing Load (FSR, DSDV, OLSR)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦.68
xiii
LIST OF ABBREVIATIONS:
AODV Ad-hoc On Demand Distance Vector
BEB Binary Exponential Backoff
DSDV Destination Sequenced Distance Vector
DSR Dynamic Source Routing
DV Distance Vector
DYMO Dynamic MANET on Demand
ERS Expanding Ring Search
FSR Fisheye State Routing
IETF Internet Engineering Task Force
LS Link State
MAC Medium Access Control
MANET Mobile Ad Hoc Network
MPR Multi Point Relay
NRL Normalized routing load
OLSR Optimized link State Routing
RD Route Discovery
RERR Route Error
RREQ Route Request
RREP Route Reply
RFC Request for Comments
RM Route Maintenance
TTL Time to Live
VANET Vehicular Ad Hoc Network
WSN Wireless Sensor Network
MANET Mobile Ad Hoc Network
WSN Wireless Sensor Network
xiv
1
Chapter 1
Introduction
2
Communication is being the integral part of every oneβs life and so are the challenges
researchers are facing. Last mile communications or infrastructure-less communication is
needed for every one belonging to any aspect of life. Such needs give birth to different kinds
of networks wired or wireless. Wireless communication being the most desired one as no mess
of installations, no worries of cable expiry or hardware constraints, freedom of movement and
more easy to use according to a user prospective.
Mobile Ad Hoc networks give such flexibility to a user and are becoming popular in market
for the said reasons. Apart from user view, it is a complex mechanism to provide
communication fulfilling such conditions. Besides many wireless communication issues, one
major issue is this that these wireless nodes/ transceivers are not always in range of each other
in accordance with their mobility. In such a scenario, there was a mechanism established that
makes each node/ transceiver a router. Hence whenever, one node wants to communicate with
a node that is out of communication range, some intermediate node between originator node
and the destined node will act as a router and make the communication between main
originator node and destined node possible.
A Mobile Ad Hoc network basically is a decentralized autonomous wireless system that has
mobility based or infrastructure-less nodes. These nodes are devices that contain transceivers
and processing capability. There are many versions of Mobile Ad Hoc network (MANET) i.e.
Vehicular Ad Hoc Networks (VANET) and Wireless Sensor Networks (WSN) are being the
most prominent ones. (IETF) Internet Engineering Task Force has defined a working group
namely MANET working group that explicitly works on routing protocols and define
standards for use within the IP suite. [1]
Several protocols have been developed on lower layers but in accordance with them, the
protocols of network layer or routing protocols play a vital role. [2]These routing protocols are
3
the main source to create, maintain and provide synchronization amongst the nodes of network
with the help of routing tables at each node of the network. Routing protocols not only have to
do the said job but they also have to cope with the ever changing positions of nodes with
different speeds, scalabilities, Transmission rates, bandwidth consumptions and many more.
Specifically, a routing protocol is said to provide efficient and quality routing.
1.1 Taxonomy of routing protocols
A routing protocol is basically a governing authority that how two communication devices
communicate with each other. More over it defines the procedures in creating/ establishing
and maintaining a route along with the data communication procedures. It is also responsible
of recovering a route in case of route failure. Numerous studies have been done and we have
categorized Mobile Ad Hoc routing protocols into major three classβs i.e.
1. Proactive Routing Protocols
2. Reactive Routing Protocols
Though now some other categories are under consideration which are geography based and
hybrid i.e. combination of reactive and proactive routing protocols for specified networks. But
still now, Reactive and Proactive routing is the prominent ones which are being studied in this
work.
ROUTING PROTOCOLS
ProactiveReactive
1. FSR2. DSDV3. OLSR
1. AODV2. DSR3. DYMO
Fig 1.1: Protocols discussed in this work
4
1.2 Proactive Routing Protocols:
These protocols are Table driven routing protocols and are meant amongst dense networks. In
such protocols, Routing information of next hop is preserved on the initialization of network
regardless of Communication Requests. As the network initializes, periodic Control packets
are flooded among nodes to uphold the paths or link states. In this way, they form a table on
each node describing the paths to/from each node. In other words, when a network initiates,
these protocols starts discovering the routes within all the nodes of network regardless of use
of that route. Such procedure may cause network over burden for a specific time but later on,
there would be no delay when a route is required within the network. in scalable and highly
dense networks that offer mobility as well, the routing information (information exchange to
keep the routing table up to date) results in huge network resources consumption. That was the
reason that reactive than hybrid and other kinds of protocols are developed. [3], [4], [5].
Features of Proactive Routing:
1. No route discovery time delay as every node has full topology routing table stored.
(Hence they provide lower latency rates for real time app.)
2. Maintaining unused or unwanted routes occupies major portion of bandwidth /
network resources, especially in highly mobile environments.
1.3 Reactive Routing Protocols:
These Routing protocols are also known as On Demand Routing protocols and their specialty
is to deal with high mobility. In such protocols, a node request for a route to its destination
only when it is needed. At nodes, routing table maintains only those routes which are in use.
5
Such routing protocols basically have a route discovery process where packets are flooded into
network in search of desired destined node. Path repairing and monitoring are being other
phases of such networks.
In both categories of routing protocols, following phases are undertaken for efficient and
quality routing.
1. Route discovery Phase
2. Route Maintenance Phase
3. Route Deletion Phase/ in reactive routing
1.3.1 Route Discovery Phase:
In a network of unknown nodes, where there is aspect of mobility is also under consideration
along with network resources usage, a protocol must be able to accomplish the route
establishment. For this purpose the initial step is to find the destined node which we want to
send data. It is also the fact that we are intended to provide two way communications not
unidirectional. For this purpose we have also to measure the route from destined node to the
source node. This process of route establishment in terms of MANET technology is known as
route discovery process.
1.3.2 Route Maintenance Process:
Once a route is established between two nodes, and data is being transferred, considering
mobility and other unpredictable aspects of Ad Hoc networks in mind, it is necessary to
continuously monitor the links using some technique. And if a link is found broken, the
protocol has to run such algorithms that can again establish the route so that the data which is
to be transmitted must be sent. This process of link monitoring and than if in case of broken
link, link repairing is termed as Route maintenance in Ad hoc Terminology.
6
1.3.3 Route Deletion:
When a route is used and data has been sent, now there is no need to store its routing
information. Routing protocol also has to delete the stale or broken or used routes to save
network resources. This process is known as route deletion process.
In our work, we focused our study to the route discovery and route maintenance phases of
reactive and proactive routing protocols for MANETs and VANETs
7
Chapter 2
Routing:
Reactive (AODV, DSR, DYMO)
&
Proactive (FSR, DSDV, OLSR)
8
2.1 Reactive Routing:
As stated earlier, route discovery initiates only when a route is required. There are numerous
protocols that are proposed and developed under this category and works well for desired
scenarios. The most widely studied, and used protocols are: Ad Hoc on Demand Distance
Vector (AODV) [16], Dynamic Source Routing (DSR) [14] and Dynamic MANET on
Demand (DYMO) [14]. We in this work are concerned only with the route discovery and
route maintenance phases of routing protocols.
2.2 Ad Hoc on Demand Distance Vector (AODV):
AODV is abbreviated by Ad Hoc On-Demand Distance vector and is reared up by classical
DSDV algorithm as described in [16]. Actually it is another flavor of distance vector routing
algorithms [1, 4], we can also conclude that it is a combination of DSDV (a proactive routing
protocol) [14] and DSR [14] algorithms. Form DSR it lends on demand features i.e. discover
paths when required, While from DSDV, it receives, loop free routes along with repairing of
link breakages. We can say that it is a modification of old DSDV algorithm that was
previously defined. Major modification is the minimization of number of unwanted broadcasts
by establishing paths/ routes on requirement basis rather than managing and maintaining full
topology routing table. The developers/ authors of AODV call it a pure on demand route
acquisition as the nodes which are not source or destination or in between the source to
destination route, are free and donβt maintain any routing table [16].
According to RFC3561 [29] AODV is designed to be used by mobile nodes of an Ad Hoc
network offering qualities of rapid adaptation of mobility of nodes, less processing and lower
memory over head. It establishes unicast paths to the destined node within the network. To
9
ensure loop freedom, a unique destination sequence number is utilized to avoid counting to
infinity or other such problems rose in previous distance vector routing protocols.
AODV, use three types of messages i.e. one for requesting a route, route request message, one
for replying a route request, route reply message and finally if a broken route is detected, route
error message (RREQ, RREP, and RERR respectively.) [4] [9] RREQ message as the name
indicates is specified for requesting a route from one node to another. When some node wants
to transmit some data to another node within the network, route is to be established first. This
RREQ message is propagated within whole network in search of a new route. Route is thought
to be determined when route request message reaches the destined node or it reaches the node
that has a fresh enough route up to the destined node. If route request message reaches the
destined node, a route replies packet is issued that follow the reverse path to the main
originating node of route request. If a route request packet reaches a node that has a fresh
enough route to destined node, than that intermediate node will generate a Gratuitous route
reply packet also termed as grat. RREP.
A route is said to be established when a RREP or grat. RREP is reached at the main
originating node of RREQ. When a route is established, link monitoring starts. At this stage,
AODV, use periodic messages (hello messages). These periodic HELLO messages are used
only by the active nodes of the route not by every node of the network; hence a great amount
of bandwidth and network usage is saved. In a route there would be numerous intermediate
nodes. During a route active time, these intermediate nodes will emit periodically hello
message to sense the topological changes occurring within the route. If a link is found broken
than after a small interval of time, a route error message is propagated from the node where
broken link is found. This RERR message shows that destined node is not reachable. This
10
reporting mechanism is based upon the concept of βprecursor listsβ [29]. These lists actually
store the address for all of the neighbors that can be used as next hop to destination.
2.2.1 Route Discovery and Route Maintenance of AODV:
In a network whenever a node intend to send some data to some other node, it needs to know
where actually the destination lies and if there is a valid route to its destination. For the said
purpose, route discovery process initiates.
Whenever a source node want to send its data to some destination node, it disseminates a
RREQ if, node needs a route to its destination. It happens when source node doesnβt have the
proper route of destination, or the route node knows is invalid or expired. In RREQ packet, a
Destination Sequence Number field also exists that presents the last known destined nodeβs
sequence number. This nodeβs own sequence number is incremented on every node the RREQ
packet passes. The Route Request identification field is incremented from the previous hop.
Before transmitting RREQ packet, the source node safeguards the route request ID and
address of source node for PATH_DISCOVERY_TIME. This is done so that when the node
receives back the RREQ packet which has already processed by it, it will not process it again
but just forward it.
After broadcasting the RREQ packet, a node has to wait for a RREP packet from destined
node. If the route destination is not found within βNET_TRAVERSEAL_TIME
(milliseconds)β, than the source node has to resent/ broadcast the RREQ packet until the
maximum of REQ_RETRIES times at maximum TTL value. At each try, the RREQ_ID is
incremented and updated.
11
The figure shows the propagation of RREQ packet across the network. To ensure loop free
and most recent path information, AODV use destination sequence number for all the routes
reaching destination from source. Sequence number and broadcast ID is maintained at each
node and on propagation of RREQ packet, each node increments its broadcast ID in the RREQ
(Route Request) packet. This incremented broadcast ID and nodeβs IP address combines to
identify uniquely every RREQ packet. Intermediate nodes can reply this route request packet
only and only if destination is in their neighborhood or have established route to the
destination.
During this RREQ forwarding process each intermediate node saves the address of neighbor
from where this route request packet is received in its routing table hence it automatically
establish a reverse path to the source. If one node received multiple copies of RREQ packet,
than it will take and forward the first copy and discards all the others.
When the route request packet reaches its destination, or reaches an intermediate node from
where a fresh route is available towards destination, than that intermediate node or destination
12
node responds by uni-casting RREP to its one hop neighbor from where it received RREQ
packet.
Normally it is expected that there is a bidirectional communications between originator and
destined nodes i.e. not only source should know the route to its destination but also the
destination to know the route to source. For this purpose, a RREP packet is generated. Only
that node can generate RREP message that itself is destination, or that has a valid route to the
destination.
If the route is discovered and RREQ packet has reached at a node which has fresh route to
destination or destined node itself than, an RREP packet is generated. Via reverse path, RREP
packet is routed back while the nodes on this route establish the forward path entries in their
routing tables pointing the neighbor from where RREP has arrived. These entries than finally
provide an active forward route to the destination from source. To avoid stale routes there is a
route timer associated with each route/ path entry. Whenever the timer expires, the route is
deleted. [8] [16]
When a route is established and link is properly periodically monitored. If during this link
sensing, routing protocol finds a broken link due to topology change or any other reason, it
13
will generate an RERR message back to main originating node of RREQ. When the
originating node receives this RERR message, it starts a new route discovery deleting the
previously stored route.
During route discovery time, i.e. when an RREQ is broadcasted the data which actually is to
be transferred from source to destination is buffered until it receives an RREP packet. If RREP
packet is received than it is transmitted on the discovered route else, it will wait for the
RREQ_RETRIES times at Maximum TTL. If even after that time, no RREP is received than
data packets are dropped.
2.2.2 Binary Exponential Back off Algorithm and AODV
Considering all the traffic on network for route discovery process, it is almost congested.
Repeated attempts by an originator node for a single destination makes the network traffic
jam. [16][13] To avoid collisions and loss of packet transmissions, binary exponential is
utilized in AODV.
Binary exponential back off is an algorithm used to space out repeated retransmissions for the
same block of data, normally used to avoid network congestion.
Binary exponential back off in AODV works in a simpler manner. For the first attempt of a
source node to search its route to destination, source node simply will wait
NET_TRAVERSAL_TIME milliseconds for RREP packet. IF Route Reply packet (RREP) is
not received during NET_TRAVERSAL_TIME milliseconds, it broadcasts a new RREQ
packet with this time it will wait for 2 * NET_TRAVERSAL_TIME milliseconds to receive a
RREP packet. For each additional broadcast of new RREQ packet, the time to wait for source
node will be multiplied by 2 from previously calculated time. This is how, binary exponential
back off algorithm works in AODV routing protocol.
14
2.2.3 BEB Algorithm for AODV RREQ is broadcasted Source node waits for NET_TRAVERSAL_TIME milliseconds= (a) If RREP reaches within NET_TRAVERSAL_TIME milliseconds= (a) Data Packets transmission initializes Else New RREQ is broadcasted Source node waits for 2 * NET_TRAVERSAL_TIME milliseconds= (b) If RREP reaches within 2 * NET_TRAVERSAL_TIME milliseconds= (b) Data Packets transmission initializes Else New RREQ is broadcasted Source node waits for 2 * (b) If RREP reaches within 2 *(b) Data packets transmission initializes
2.2.4 Expanding Ring Search Algorithm and AODV
The binary exponential back off algorithm in AODV has limited the generation of timely route
requests and congestion of an AODV network. A simple approach for broadcasting RREQ in
AODV can be of blind flooding technique that is each node has to rebroadcast the packet
whenever it receives it for the first time.[16] This is an easy way of broadcasting but this way
can cause BROADCAST STORM PROBLEM by generating redundant transmissions. To
avoid this broadcast storm problem, there are many optimization techniques and one of which
is expanding ring search algorithm. In this mechanism, the flooding is controlled via TTL
value though ERS also uses blind flooding for broadcasting.
15
In ERS, initially on broadcasting of source node,[16] TTL value is set as 1. If RREP is not
received during a specified time, than RREQ packet is again broadcasted by incrementing
TTL value by 1 and waiting time is doubled from the previously calculated waiting time. This
process continues till the destination is discovered or it reaches the last limit of expanding ring
search broadcast. The figure produced by [16] very clearly explain the expanding ring search
algorithm.
2.2.5 ERS Algorithm for AODV
According to AODV RFC 3561, ERS is used under following parameters
Originator node initially uses TTL= TTL_START in RREQ packet IP header RREP Time out= (RING_TRAVERSAL_TIME=2 * NODE_TRAVERSAL_TIME *(TTL_VALUE + TIMEOUT_BUFFER)) If RREQ times out and no RREP is received, the source node broadcast the RREQ again while TTL= TTL_INCREMENT. The same process continues unless, RREP is received by source node or TTL value reaches TTL_THRESHOLD Beyond TTL_THRESHOLD, TTL= NET_DIAMETER is used for each attempt.
16
When TTL=NET_DIAMETER, the timeout for waiting RREP is set to NET_TRAVERSAL_TIME(NET_TRAVERSAL_TIME= 2 * NODE_TRAVERSAL_TIME * NET_DIAMETER) 2.3 Dynamic Source Routing (DSR):
DSR abbreviated by dynamic source routing protocol [14] is basically on demand routing
algorithm designed upon the concept of source routing. Each node in the network carries a
cache that saves the source routes. The updates in these caches are continues as new routes are
known. DSR is also stated as first on- demand source routing based protocol [5, 6]. Source
routing basically allow the sender of a packet to partially or completely specify the route by
which the packet is traversed through the network. The path is based upon destination of the
packet.
2.3.1 Route Discovery and Route Maintenance of DSR:
In DSR, some extra route discovery features are adopted i.e. caching Overhead routing
information, replying to route requests using cached routes and route request hop limits. [14]
DSR specifies that any node which is forwarding or just hearing any packet should store all
the routing information from that packet which can be useful in its own cache. RFC4728
termed it as CACHING OVERHEAD ROUTING INFORMATION. RFC4728 also specifies
that any node that receives a RREQ where this node is not the destined node, it will searches
its own routing cache for the route towards destined node of the RREQ. If a route to destined
node is found in the route cache, this node will not forward this RREQ message further,
instead it will generate route reply back to the main source node. Such route reply is termed as
βgrat. RREPβ. DSR also adds a HOP LIMIT field in the header of RREQ message. This hop
limit field can be used to limit the number of hops up to which, route request has to be
flooded. It is an easier way to implement expanding ring search algorithm. This hop limit
17
implements in accordance with the βTime To Liveβ TTL field in the header of RREQ message.
As this message is propagated, hop limit decrements, and route request packet is expired when
hop limit reaches zero value provided if destination is not found.
DSR uses Source routing in which packet has ordered list of nodes in its header by which this
packet has traversed. [7] In this way, source node or sender selects and take control of the
routes used by and for its own messages.
When some node wants to transmit data to a destination in the network, initially, it queries
whether the route exists from source node to destination node in cache/ buffer or not. If there
is some route stored from source to destination in buffer/ cache, than it sends data else route
discovery process is initiated.
If no route is found in the cache, the source node floods RREQ packet in the network. When
an intermediate node receives a RREQ packet, it checks if this packet has reached again or
not. If node found repetition, it will discard that RREQ packet else it will attach its address to
the route record in the head part of packet. After that, intermediate node will forward it to all
of its adjacent nodes.
18
When destination node receives the RREQ packet, it reverses the route records in route
request packet and returns back a RREP packet on the same reversed route. Source node when
receive the RREP packet it immediately saves the route information in its cache for future use.
When a route is established between source node and the destined node, than link monitoring
activates. In DSR, advanced maintenance concepts are introduced as packet salvaging; queued
packets destined over a broken link, automatic route shortening and increased spreading of
route error messages [14].
The concept of packet salvaging specifies that when a node which is taking part in forwarding
packets during data transmission session detects via periodic acknowledgements that next link
is broken. If node have an alternate route saved in its cache, than it will not discard the packet
rather salvage it. Packet salvaging actually refers the replacement of original routing data in
the packet than alternate data which was saved in the cache of node as alternate. The packets
will then be stored within the node. Besides packet salvaging, when an active node (node
participating in a link) detects the nest link is broken, it will intend to salvage the packet. If
19
there is no alternate route saved in the route cache, than the packets are to be buffered at
nodeβs network interface queue and maintenance buffer [14].
Automatic Route Shortening is also done in DSR in the same way as in [14]. The concept
behind is this that there may exist some unnecessary intermediate nodes that can be bypassed
without effecting the route. As in DSR, a node hears and saves useful information from the
packets passing through neighboring nodes, and it get to know that some of any one of the
intermediate node can be bypassed than, it simply deflects the route, shortening it. Nodes need
to be in promiscuous mode for such operations.
Taking an very simple example of link monitoring and link breaking, suppose that there are 3
intermediate nodes βxβ, βyβ and βzβ in between the source and destination nodes βaβ and
βbβ respectively. Hence the links of the route are a-x, x-y, y-z and z-b. Now according to [14]
βaβ node is responsible for the link (a-x). Hence node βaβ require periodic acknowledgments
from node βxβ in response of the acknowledgement requests from node βaβ. If during a
specified interval of time or specified number of acknowledgement request attempts, no
acknowledgements are received, the link is considered as broken. On discovery of broken link,
a RERR message is generated. The broken route is deleted from memory and route error
message is than transmitted to main source node via reverse path.
2.3.2 Binary exponential Back OFF and DSR:
If a node retires many times to search a route for a single destination and in the same way
some other nodes at the same time are requesting for routes to their destinations, there would
be large number of unproductive route request packets that almost congest the network.
[14]To avoid such jumble up of route request messages, binary exponential back off algorithm
20
is used just as in AODV. For every failure, it simply doubles the time to resend or rebroadcast
the same route request.
2.3.3 Expanding ring search Algorithm and DSR:
RREQ packet in DSR also contains a hop count field that represents number of hops of any
described route. If hop limit field is set to 1 than only its neighbors will receive the packet and
they will not further broadcast it as the hop limit was 1 and it has been full filled. If no route
reply packet is received by the source node at the hop limit of 1, it will increment the hope
limit and rebroadcast the packet. [14] This hop limit is implemented using TTL field in the IP
header of the packet carrying the route request. As this request passes from node, its TTL field
is decremented and the route request packet is dropped when that TTL field reaches at zero
before finding its target.
ERS approach could increase the average latency of the route discovery as multiple discovery
attempts and timeouts may be needed before discovering a route i.e. from source to
destination.
2.4 Dynamic MANET on Demand (DYMO):
DYMO stands for Dynamic MANET on Demand is another proposed reactive routing
protocol and is meant for highly mobile nodes in multi hop networks. It is forecasted that such
mobile networks are future networks. In 2005 DYMO draft was submitted [11] [29] it attained
lots of attention and this is the reason that draft has been modified many times. In some
manner we can say that DYMO base upon reactive routing protocols like AODV and DSR.
Main features of DYMO routing protocol is path discovery and path management. This is the
cause that DYMO is not so simple and easy to implement but on the other hand this protocol
21
is flexible, scalable and bears other enhancements in capability and expansion [12] [14].
Moreover, DYMO routing protocol can work either on IPV4 or IPV6 hence we can
interconnect network with internet as well.
Just like AODV, this protocol also works on RREQ and RREP packets for path discovery
process. A RERR packet is also used in this routing protocol which is responsible of error
notification within some established route.
2.4.1 Route Discovery and Maintenance of DYMO:
The source node, in DYMO routing protocol, like AODV, broadcast RREQ packet in the
entire network to reach destination node [15].
Once a node requires discovering a route, it initiates dissemination of RREQ message through
the entire network. During this hop-hop process, each intermediate node keeps the record of
route to the main originator node of RREQ. When a destined node receives a RREQ, like the
procedure of reactive routing protocols, it sends a reply message RREP to confirm and
establish a route. During the hop β hop dissemination of RREP message, again this time the
22
intermediate nodes saves the route to destined node in their routing tables. When the main
originator of RREQ message receives the route reply message, the route is said to be
established [29].
Route maintenance in DYMO is divided into two operations. In order to preserve active
routes, DYMO extends the route life time if a packet is successfully transferred via that route.
And 2ndly, each route is monitored and if in case one link of a route breaks, than as in he
basic theme of reactive routing protocols, route error message is initiated. When the main
originator node of RREQ receives an RERR message, it initially deletes the route and then
new route discovery process is initiated.
Accuracy and security is maintained by digital signatures and hash functions are also used in
this routing protocol. For the discovery of new route, the originator DYMO node broadcast
RREQ message in the network to find its destination. [10]During this hop by hop searching
procedure, each intermediate node saves the route of source or originator node.
When the destination node receives RREQ packet, it issues an RREP message/ packet that is
sent to originator or source node hop by hop. Each intermediate node at this time as well
23
records the route to destination. When the RREP packet is received by the source node, route
is established between originator and destination in both directions.
During this broadcasting each intermediate node saves a route to the source node. When the
destination node gets the RREQ packet, it than responds to the RREP packet, unicasting it
towards the source node by using reversal of saved paths. On each iteration, the intermediate
nodes save the route to the destination till the RREP packet reaches source node.
2.4.2 BEB and ERS in DYMO
Like AODV, when a node is required to search a route, it broadcast a route request packet
RREQ to find a valid route to some destination. The source node than waits for a unit time to
get the route reply but if reply is not obtained than it again rebroadcast the RREQ packet. To
avoid the repeated attempt that makes a network congested, a binary exponential back off
algorithm is used. The data packets which are waiting for a valid route to be transmitted are
kept in a buffer of fixed size. If route discovery is attempted at its maximum times and yet no
destination is found than, data packets for that undiscovered destination are discarded or
dropped from buffer and an ICMP, un reachable destination packet is delivered. If route was
discovered but link has broken than a route error message is issued.
Binary exponential back off functionality in DYMO can be expressed with the following
procedures.
2.4.3 BEB Algorithm for DYMO:
RREQ packet is broadcasted
Source node waits RREQ_WAIT_TIME for a route to destination
If route not found within RREQ_WAIT_TIME send another route request packet.
Source then has to wait for 2* RREQ_WAIT_TIME.
24
If even than route not found
Rebroadcast RREQ and wait for 2*(2* RREQ_WAIT_TIME)
This procedure continues and at every rebroadcast the source node has to wait for route twice the previous time
If route is not discovered until DISCOVARY_ATTEMPTS_MAX limit reaches.
AT this level, route discovery process is assumed as failed
Data packets at source node waiting to be transmitted are buffered.
BUFFER_SIZE_PACKETS or BUFFER_SIZE_BYTES stores a fixed and limited size of data to be buffered.
This buffer act as FIFO (first in fist out). The packet that is buffered earlier will be removed earlier.
Buffer setting BUFFER_DURING_DISCOVERY should be manually or intelligently controlled as it is configurable.
After DISCOVERY_ATTEMPTS_MAX, without receiving the route, destination unreachable ICMP packet is issued and data packets in buffer for the said destination are dropped.
During all this process, to control the RREQ propagation network wide and to reduce the over
head, DYMO may use expanding ring search algorithm. In DYMO routing protocol, ERS
algorithm can be implemented at MsgHdr.Hoplimit in RREQ packet along with all the
procedures, AODV routing protocol adopted.
There is a very small difference in the route discovery process of AODV and DYMO. AODV,
only generate routes for the destination whereas, DYMO, at each intermediate node towards
the destination store routes in its routing table.
25
2.5 Comparison: ERS and BEB in AODV, DSR, and DYMO:
Flooding is an essential part of AdHoc networks especially in the process of route discovery.
On the other hand, this is the major cause of network overflow. The route discovery messages
can generate a broadcast storm problem in such networks. To avoid or to control this problem
several techniques and algorithms are presented and implemented. Binary exponential back
off, expanding ring search, Directed flooding and Query localization are the algorithms that
are used in AODV, DSR, DYMO and TORA routing protocols.
In AODV, DSR and DYME, ERS is used to avoid or solve broadcast problem in any network
while BEB is utilized to avoid collisions and loss of data traffic along with being part of
network congestion control mechanism. The Mechanism of Expanding ring search Algorithm
in AODV and DYMO are same as DYMO adopted the same procedure as that of AODV.
AODV , DYMO and DSR use Both of these techniques in simultaneously but with a little
difference in parameters and achieve the said goals of these mechanisms. In all these routing
protocols, expanding rings circles algorithm is utilized with the help of TTL field in header of
the packet but with a little difference.
In AODV and DYMO, the TTL field is based upon time factor, while in DRS, TTL field is
based upon hop count factor.
The rings made in AODV are incrementing that unit time which normally is set as
RING_TRAVERSAL_TIME milliseconds and each time, when route request packet fails to
reach its destination in the ring and no route reply packet is received by source node in given
time period, that RING_TRAVERSAL _TIME milliseconds is incremented and new route
request packet is broadcasted with extending ring via more time for the route request packet to
find its destination.
26
In DRS, the mechanism though is implemented via TTL field as well but in procedure it is a
bit simpler than that of AODV adopted. In DRS, the TTL field is dependent of hop limit field.
If hop limit field is set to 1 it means that the route request packet will be dropped or discarded
when the one hop search is done. In this way, the hop limit field can be modified and new
route request packet can be sent to find the destination with in that hop limit. Suppose that
hope limit set is 10 than at each hop, the hop limit value is decremented and when hop limit
reaches its value at 0, the packet is discarded.
In AODV, the processing for TTL value is much greater than that of DRS.
If we consider DYMO, it also works almost similar than that of AODV. Its binary back off
exponential time is same as that of AODV and DSR i.e. multiple of 2 on each iteration and it
copies exactly the same algorithm for expanding ring search as used in AODV. The only
difference between AODV and DYMO is this that AODV, only generate routes for the
destination whereas, DYMO, at each intermediate node towards the destination store routes in
its routing table.
In DYMO routing protocol, the information of routes is recorded at intermediate nodes with
the help of RREQ and RREP packets while in DSR, the routes are stored in the header of
packet.
These three routing protocols use ERS and BEB for controlling the broadcast storm problem
but if we consider another routing protocol like TORA, it doesnβt use these algorithms.
Besides it uses the concept of directed flooding and query localization. Directed flooding is
made possible with the help of Directed Acyclic Graph which limit the broadcasting in one
direction i.e. upstream to downstream, form source to destination while query localization
technique works well with the topological changes and partitioning of network.
27
2.6 Proactive Routing:
As discussed in previous chapter, whenever a network with proactive routing protocols
initiates, the route discovery for every possible destination also initiates. Proactive routing
protocols donβt wait for the data to be transmitted and then search only required route as in
reactive routing protocols. They initialize and maintain the routing table at each node. That is
the reason that latency rate is much lower in proactive routing protocols than that of reactive
routing protocols. But on other hand, proactive routing protocols raise the issue of network
resources due to flooding nature. Widely studied and practiced proactive routing protocols are
Fisheye State Routing (FSR), Destination Sequenced Distance Vector (DSDV) and Optimized
Link State Routing (OLSR) protocols.
2.7 Fisheye State Routing (FSR)
In proactive routing protocols, there is a huge problem of initial flooding of packets for
topology and routing information. This process may choke the network or use maximum of its
resources. To solve this problem, Fisheye state routing algorithms defined the phrase βMulti
level scopeβ that is able to reduce the routing overhead in dense networks. Fisheye state
routing floods the periodic control messages of a destined node with respect to its neighbors
with a frequency that is dependent on the hop distance to that destined node. As the
correspondent of these control or status update messages moves away from the destination, the
information propagated gradually declines to lower frequencies. From such status updates or
control messages, every node in the network built and maintains a routing table. This routing
table is precise for the nodes nearby but as the hop distance increases; the routing information
in the same proportion fades or gets imprecise. Hence the route on which a packet travels may
seems to be faded but as it gets closer to the destination, route becomes more precise and
accurate. FSR follows the link state algorithms as it issue periodical updates of link state, but
28
instead of flooding these periodic updates to whole of the network, it floods in step wise
manner. In this way, FSR results in huge savings of routing overhead which was caused in
maintaining the topology details of every node in the routing table of every node. Fisheye state
routing is briefly discussed and implemented by [20].
Fish eye state routing [1, 2] is a routing protocol providing a tree like structure. It updates the
link state information in different frequencies that depends upon fish eye scope distance. That
is frequencies are higher for nearer nodes and lower for far away nodes. Within scalable and
dynamic environment, a packet as reaches near the destination, the routing gets more accurate
regardless of mobility and scalability. FSR basically provide, simplicity of routing, gives
updated shortest routes , provide robustness in mobile and scalable environments and one of
the major benefits is the reduction of routing overhead [19].
2.7.1 Fisheye Scope Technique:
FSR use fisheye technique given by [18]. According to a fish eye structure, it can capture the
details of nearby things but fades for far away things. [17] In routing the same approach is
utilized that within neighbors of a node, precise information is gathered but as the hop
distance increases, the details get fader.
Fisheye scope technique basically permits exchanging routing information at different
intervals for the nodes existing at different fish eye scope or fish eye distances. This gives a
vital effect on routing overhead by reducing the size of the message. According to [19], 37
Fisheye state routing broadcasts the topology messages only to the neighboring nodes. With
such optimizations FSR proves to handle scalable and dense networks.
Fisheye state routing, routes every packet In accordance with local computing, i.e. each packet
is routed with respect to the routing table of the node it is passing through. The routing table of
29
the node by which packet is passing through is handling the most up to date routing
information and as the basic theme of FSR, as the packet gets nearer to destination, the routing
becomes more precise. 37 For inner scope of fisheye, the routing is accurate due to short
intervals but for outer scope the routing information may be outdated due to longer intervals.
If we consider a dynamic and highly mobile environment, the accuracy of outer fisheye scopes
will be more blur and inaccurate but on the other side, when the packet reaches a node that has
destination in its inner fisheye scope; the routing information again gets precise.
Fisheye scope routing donβt use any trigger [19] for repairing a link or when a link is found
broken, it simple will not include this link in its next fisheye scope control message exchange.
Whereas FSR use sequence numbers along with table refreshment to remain updated with loop
free routing table.
Maximum accuracy in scope 1
Minimum accuracy in Last
Scope
SCOPE 1
SCOPE 3 SCOPE 2
Fig 2.7:
According to [19] the fisheye scope is represented as the nodes that are accessible within some
limited number of hops. Fisheye scope can be better understandable by the above illustration.
30
2.7.2 FSR WORKING:
In a network, where Fisheye state routing is operating, each node will maintain the following
tables. On the basis of these tables routing is made possible.
Topology Table:
This table maintains the information contained by periodic link state messages. Routing table
is made on the basis of this table that contains the LS information and destination sequence
number. Each topology table includes
1. Destination address field
2. Destination sequence number field
3. Lint state list
Neighbor link state table:
On receiving a LS message, link state information of the sender node is saved in neighbor link
state list. If a node didnβt get any LS message from some neighbor for a
NEIGHBOR_TIMEOUT period, that neighbor will be deleted from nodeβs neighbor link state
table.
Routing Table:
Routing table comprises of following fields
1. Destination Address
2. Next hop Address
3. Distance
31
Link Breakage:
Ad Hoc networks are known for mobility and mobility causes link breakages. In Fisheye state
routing each node uses soft state to find a broken link. i.e. as according to neighbor link state
table, a node will delete all the routing information from that neighbor if it appears not in
contact for a NEIGHBOR_TIMEOUT time value.
2.8 Destination Sequenced Distance Vector (DSDV)
DSDV is one of the parent Ad Hoc networks routing protocols. Basically it is derieved from
Bellman Ford algorithm [21] [22]. Bellman ford Algorithm basically provides solutions for
shortest path between source node and destined node [6]. DSDV in addition of classical
Bellman Ford Algorithm introduces a new feature i.e. sequence number for each routing table
entry of whole of the network. In DSDV, Routing Table on each node make lists of all the
possible destined nodes within the underlying network along with their number of hops and a
sequence number to prioritize the routes. This routing information is broadcasted or multi
casted to the neighbors. Not only periodically the change in topology is also transmitted in
form of triggered updates whenever a change in topology due to mobility or any other reason
occurs. C Perkins presented this protocol in 1994[6]. There is some detailed description of
protocol provided in [6] Till to-date, numerous comparisons have been made between DSDV
and other routing protocols both reactive and proactive in nature.
2.8.1 Route Discovery and Route Maintenance of DSDV:
DSDV transmits control packets periodically for updating of topological changes or when a
link breaks it issue a trigger message. This aptitude of protocol makes the network suffer with
delay in updating of routing information when the environment is highly mobile or dynamic.
There are two types of updates considering DSDV.
32
1. Incremental Updates
2. Routing table updates.
The new featureβ sequence numbersβ are utilized to ensure loop freedom and attaining
information about shortest possible route to destination. Any route that has the highest
sequence number is prioritized among other routes and if the sequence number for two routes
having same source and destined nodes are equal than decision is to be made on lower metrics.
The one route that has lower value of metrics must be used. According to [22], [6] the routes
which are broken normally use even numbered sequence numbers. This assumption also helps
in loop freedom.
DSDV use two types of routing tables.
1. Forwarding Table:
i. The forwarding table contains the information addresses of all nodes
within the underlying network
2. Broadcast table:
i. The broadcast table calculated the time for periodic updates or
broadcasts advertisements
33
All the routing information, routing info or periodic updates uses these two tables. Routing
table of a node mainly comprises of
a) Nodes address,
b) Next hop address
c) Routing metrics
d) Destined node sequence number
As the basic theme of proactive routing, all nodes with underlying DSDV protocol maintain a
routing table for the interconnectivity of different nodes. These routing tables are maintained
for all the available destinations and number of intermediate hops to the destination. Each
destination route in the routing table is distinguished with a unique sequence number. To
remain interactive and gather latest routing information, periodic control messages are
transmitted. And with the help of these periodic control messages, the routing tables of every
node are updated. According to [6] each node that broadcasts the data, will contain its new
sequence number along with the
a) Destined node address
b) Hop count to reach destined node
c) Destined node sequence number
The routing tables that are shared also have the hardware address of mobile node forwarding
them. They will also contain sequence number of source node. The latest destination sequence
number is prioritized on making packet forwarding decisions. When every a new sequence
number is generated, it is updated to all nodes of the network. This new sequence number is
used in maintaining and updating the routing table entries of source node. When route
information is received by a node, it simply advertises this information among all the
corresponding nodes. In case a broken link is found than the protocol donβt wait for the next
34
periodic control message to flood this broken link information but issue a trigger message on
the vary time.
Though DSDV provides many benefits like loop freedom, infinity count problems, and
avoiding extra traffic using incremental updates etc but in contrast to these plus points, DSDV
also make the underlying network suffer with the bandwidth wastage due to unwanted routing
advertisement, not supporting multipath routing and not efficient in larger/ dense networks.
2.9 Optimized Link State Routing (OLSR):
Optimized link state routing protocol is a proactive routing protocol for MANETs [24], [25]. It
follows the basic concept of link state routing along with the Multipoint relay concept. Each
node in OLSR routing protocol selects a set of multipoint relay nodes which are its neighbors.
Only MPRs forward the control packets in such a way that information should reach whole of
the network [24]. These selected MPR nodes are held responsible for declaring LS
information in entire network. Multipoint relays are also used in route calculations from a
source node to destined node. An MPR selected from a node must have a symmetric or bi
35
directional link to minimize the problems of packet transmissions over asymmetric or
unidirectional links. Basically in classical link state routing, only two modifications /
optimizations are made to make optimized link state routing protocol [26]. The concept of
MPR set of nodes that are responsible of broadcasting topology control messages. And 2ndly
contents of topology control message are reduced.
2.9.1 Multipoint Relay Concept:
The vital role of minimizing routing overhead in optimized link state routing protocol goes to
the multipoint relaying concept. [23] This idea was generated to reduce the routing overhead
by minimizing the retransmissions of the same packet in the same region. This concept limits
the number of retransmissions by limiting the number of re-transmitters. It is achieved by
selecting some nodes that belongs to one hop neighbor-hood of a node instead of all neighbor
36
nodes of that node. The neighboring nodes that are selected as multipoint relay set of a node
are permitted to rebroadcast the packet they receive while rest of the nodes which are not
selected as multipoint relays will only receive the packets but wont forward them. [RFC
OLSR] Only nodes that have maximum coverage to 2 hop symmetric neighbors of the node
and by self they are one hop symmetric neighbors of that node can be selected as Multipoint
relays of said node. In order to achieve this, link state information is a must in between the
nodes. This Link State information is provided using periodic Hello Message.
2.9.2 Route Discovery and Maintenance of OLSR:
According to RFC 3626, OLSRβs functionality is divided into two aspects, i.e. the core
functionality and the Auxiliary functionality.
RFC 3626 specifies the core functionality of OLSR as: packet forwarding, packet formatting,
periodic hello messages, link sensing, neighbor detection, topology discovery, calculating
routing table and node configurations.
In Auxiliary functionality, the protocol deals with: non OLSR interfaces, notifications form
layer 2, redundant MPR flooding etc.
There are 4 types of messages utilized in OLSR.
1. Hello messages:
i. Used for neighborhood detection
ii. Selection of MPR set
iii. Link Sensing
2. Topology control Messages(TC)
i. Route Calculation
3. Multiple interface Declaration messages(MID)
37
i. An OLSR running node having more than one interface, than relation between main address and other OLSR interface addresses is exchange via MID messages.
4. Host and Network Association messages(HNA)
i. Allows external routing information in OLSR underlying network.
When a network initiates, Periodic Hello messages are broadcasted for the purposes of
neighbor detection. Once the neighbors are detected, MPR selection initiates. For this purpose
MPR selectors are made that selects the MPR set of nodes of a specified node. When MPR set
of nodes is selected by MPR selectors, The Topology control messages are broadcasted to
entire network. These messages actually calculate the route and spread this route calculation to
every node of the network. Only selected MPRs will flood the TC messages hence saving lots
of network resources. If the number of MPRs in a network is low, the routing overhead will
also be lower. Once the route information is distributed amongst all the nodes, there is another
feature of OLSR. It offers multiple interfaces to a node. These interfaces may be OLSR
interfaces or interfaces for other protocols. One on these OLSR interfaces is termed as main
address of the node on which major transactions will happen. But at times, other OLSR
interfaces may also be utilized. To exchange the relation between main address and OLSR
other interface addresses, MID messages are spread. As discussed earlier, that OLSR also
provide interfaces to other networks as well. And for that purpose, HNA messages are
broadcasted periodically.
38
Chapter 3
Analytical Modeling:
Reactive & Proactive
39
3.1 Reactive Modeling
Unlike proactive routing protocols, where all the routes are formulated whenever the network
initializes, in reactive approach, routes are queried only when needed by a node. A route
request is flooded in the entire network and when a route is established, data is to be sent.
Route Discovery and route maintenance are the two major aspects of routing overhead of a
reactive routing protocol.
3.1.1 Route Discovery:
When a node wants to transmit some data to another node, and there is no valid route
stored in its routing table, it will flood the RREQ packet. An RREQ packet comprises of many
fields, most prominent ones are Source identifier field to identify the route requesting node,
Destination identifier field to identity the destination node and TTL field to limit the flooding
or other purposes that can be defined according to need of the protocol. This uniquely
identified RREQ is flooded amongst all the nodes of network until it reaches destination node
via different paths/ routes. Via RREQ packet which has reached destination node, destination
node will keep the route back to source and send an RREP packet to all those routes from
which RREQ has reached it. On every node, where packet reaches, hop count / TTL is
incremented / decremented and route table entries are updated. [30]
3.1.2 Route Maintenance:
Whenever a route is established via RREQ and RREP messages, link sensing initiates with the
help of periodic messages [29] a link can be deteriorate due to noise or topology change. This
is the main reason that link is being monitored periodically. In either case when a node finds
no link to next hop, it issues an RERR packet informing the un-reachability of destination
40
node that is transmitted back to main source node. On receiving a RERR packet, the main
source node initiates new route request for broken link. [31]
In this work, we are concerned only with the basic architecture of route discovery and link
monitoring phase of reactive routing protocols. Generally following are the steps for Route
Discovery and Route Maintenance involved in almost all of the proposed reactive routing
protocols. These two phases of a routing protocol gives the routing overhead of that protocol.
3.1.3 Steps involved in Route Discovery and Route Maintenance:
1. Flooding RREQ packet (route request packet)
2. Receiving RREP packet (route reply packet) when destination node is found by RREQ
3. Link is established and now link monitoring starts by periodic messages.
4. When the link is found broken, different methods apply to rectify this problem.
5. New route discovery / local repair/ wait for time out happen for the link that has broken.
Form the above mentioned 5 steps of Route Discovery and Route maintenance; we have
analytically modeled first three steps of main architecture of Reactive Routing protocols.
In this work, we have analyzed two types of scenarios i.e. the one where there is only one link
active in the network and a source node S wants to create a link to its destination node D
during network life time T. And in other case we have tested the limits of a network of n
nodes where every node is eager to send its data during network life time βTβ.
3.1.4 Routing overhead for a single Route:
Routing over head of a single link is the sum of broadcasting of Route request (RREQ)
packets to all the nodes of network and the Route Reply (RREP) packets received by the
source node
41
(We considered that a single link between a source node S and destination node D can have
multiple routes.)
It can be written as
π π(π) = π π[π π·π] + π π[πΏππ]
Or in some detail
π π(π) = [π π πΈππ] + [π π πΈππ] + π π[πΏππ]
Where
RO(i) = Routing over head of one link
RO[RDi] = Routing overhead of Route discovery phase of one link
RO[LMi]= Routing overhead of Link monitoring of one route
[RREQi] = Expected route requests for one link
[RREPi] = Expected route Replies for one destination to one source
T = Route Life Time
t = periodic update interval
3.1.5 Overhead due to Route Request packets:
Assumptions:
n= number of nodes in a network
A= area of network = (X x Y) sq. m.
42
r= Communication range of node= π r2
Considering the above suppositions, we can say that if a node floods a packet than, that packet
will reach (n-1) nodes of the network. As discussed in [27] the any node in the network must
has some neighbors. Average number of neighbors in network for any node can be defined as
Navg = (π β 1) ππ2/π΄β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦....1
Navg is the average number of neighbors of any node.
The Route Request packet when flooded, than routing overhead due to RREQ packet at 1st hop
would be equal to number of neighbors of packet initiating node. Hence it can be written as
[π π πΈπ1π π‘ βππ ] = (π β 1) ππ2/π΄
The number of route request packets for one route can be calculated as
43
[π π πΈππ] = β (π β 1) ππ2/π΄ππ=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦2
3.1.6 Overhead Due to Route Reply:
When a route request packet (RREQ) reaches its destination, destination will generate Route
Reply Packet (RREP). This packet has to reach the main source node in order to establish a
route. In RREQ packet header, there is a field of hop count that stores the number of hops this
RREQ packet has passed to reach destination. Now the destination knows that how many hops
are there in between destination node and source node.
Mover over according to Route replies packets follow the reverse path and there may be a
number of routes that are found from one destination. Considering these all aspects, we can
conclude that overhead due to Route Reply packets for one route can be expressed as:
[π π πΈππ] = β π» (π β 1) ππ2/π΄ππ=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦β¦β¦3
Where
44
l = number of routes discovered by RREQ packet for a single destination
H= number of hops of one route.
Route Discovery overhead can be expressed as
π π[π π·π] = [π π πΈππ] + [π π πΈππ]
π π[π π·π] = β (π β 1) ππ2/π΄ππ=1 + β π» (π β 1) ππ2/π΄π
π=1 ......................................................4
3.1.7 Monitoring over head of one link:
When the route is established, link sensing or monitoring starts. According to [29] when link
is established periodic hello messages are used for continues link monitoring. As discussed
earlier the route life time isβ Tβ whereas periodic update interval time is βtβ. Than number of
periodic messages propagated during the network life time would be βT/tβ.
Suppose that there are βmβ nodes in a route. Hence m nodes will broadcast hello messages for
link sensing. Average number of neighbors of any node is discussed in eq. 1 than, routing
overhead due to link monitoring of one route:
π π[πΏππ] = ππ‘β ((π β 1) ππ
2
π΄ππ=1 )β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..5
45
Over all routing overhead due to route discovery and route monitoring (for one route) can be
expressed as:
π π(π) = β (π β 1) ππ2/π΄ππ=1 + β π» (π β 1) ππ
2
π΄+ π
π‘β (π β 1) ππ
2
π΄ππ=1
ππ=1 β¦β¦β¦β¦β¦β¦β¦6
3.1.8 Routing overhead for n Routes:
Routing over head for n routes can be calculated as
π π(π) = β [π π πΈππ]ππ=1 + β [π π πΈππ]π
π=1 + β [πΏππ]ππ=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.7
Placing values:
π π(π) = β β (π β 1) ππ2/π΄ππ=1
ππ=1 + β β π» (π β 1) ππ2/π΄π
π=1ππ=1 + β π
π‘β (π βππ=1
ππ=1
1) ππ2
π΄β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.β¦.8
46
Now we have two optimum functions with respect to route discovery and route monitoring
mechanism of reactive routing i.e. f(x) and f (y). f(x) represents the routing overhead of route
discovery and route monitoring of a route while f(y) represents route discovery and route
monitoring of n routes. If we want to analyze the routing protocol behavior with respect to
changes in range of a node, route life time, area of network and periodic update interval times,
we have the optimum function x(r, A, T, t). Taking partial derivatives of f(x) and we get
Ξ΄x/Ξ΄r = β (π β 1) 2ππ/π΄ππ=1 + β π» (π β 1) 2ππ
π΄+ π
π‘β (π β 1) 2ππ
π΄ππ=1
ππ=1 β¦β¦β¦β¦β¦β¦9
rate of change of size of network will effect as equation 10
Ξ΄xΞ΄A
= β (π β 1) ππ2
βπ΄2ππ=1 + β π» (π β 1) ππ2
βπ΄2+ π
π‘β (π β 1) ππ2
βπ΄2ππ=1
ππ=1 β¦β¦β¦..β¦10
Partial derivative of T will be:
Ξ΄x/Ξ΄T = 1π‘β ((π β 1) ππ
2
π΄ππ=1 ) β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.11
if we want to analyze overhead on increased the periodic update interval, we will get the
following equation
Ξ΄x/Ξ΄t = πβπ‘2
β ((π β 1) ππ2
π΄ππ=1 ) β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..β¦β¦β¦12
Equation 9 validates that if the communication range of nodes increases; there would be a
decrease in routing over head.
Equation 10 expresses that increasing the Area of network will also increase in routing
overhead as number of hops and number of neighbors are nonlinearly concerned with the area
of network.
47
Equation 11 shows that if we increase the route life time, overhead will increase.
Equation 12 shows that increase in the periodic interval time will result in decrease of routing
overhead.
3.2 Proactive Routing:
As we know that in proactive approach, whenever a network initializes, all the routes are
created immediately using flooding mechanism. Further onwards, that complete routing table
is updated periodically with the help of periodic messages. If any change occurs between two
periodic messages, a trigger message is broadcasted as described in DSDV [6] (a proactive
routing protocol) instead of waiting for next periodic update message. In a dynamic network,
there may be loss of packets due to broken links which are not updated at that moment. Hence
if we consider routing over head of a proactive routing protocol, than we can say that it is the
sum of number of packets failed due to link breakage, periodic messages that are broadcasted
after every specified instance of time and triggered update messages due to change in topology
in between periodic updates.
π π = ππΉ + ππ + ππ
Where
RO= Total routing over head of protocol
PF= packets failed to reach node due to link breakage
PR= Periodic update messages
TR= Triggered update messages
48
Normally there are two types of errors that lead to packet failure as discussed in detail in [32]
one error is this that, a node knows that it has a neighbor X where as node is unable to
communicate it where as the other error is this that a node donβt know that there is a neighbor
X from which it can communicate. In either case, the probability of packet loss is increased.
3.2.1 Route failure impact on routing Over head:
During periodic update time span (Tpr), number of packets encountering route failure is
defined in [33]
ππ»(ππΉ) = οΏ½ οΏ½ οΏ½Qrl (Tpr)Na(TPR)οΏ½L1
π=0ππβππ΄β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦13
Where
OH(PF) = Overhead of packet failures due to link breakage.
Qrl (Tpr) = probability that during first r hopes, the uplink state will not change its state
to down link.
Na (Tpr) = Number of Data packets arriving at time Tpr
Tpr = Periodic route update time
Li = length of Pi (ith Path)
PA = set of all paths in the network
3.2.2 Periodic Message Overhead:
Periodic message over head in proactive routing protocols can be stated as size of routing table
per periodic route update time. While routing table size is equivalent to the size of network.
49
Combining it with the complexity of routing over head we get the periodic message update as
discussed in [33]
ππ»(ππ) = πΎπ3
BβTprβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..14
Where
OH(Pr)= over head due to periodic updates
B = Bandwidth
n = Number of nodes in a network
K = used to adjust routing protocol impulse factor
3.2.3 Trigger Overhead
Coming to the third important aspect of overall routing over head, the triggered messages, we
have to understand when and how actually a trigger message is updated.
Consider that there is a node in a network that moves in such a way that it changes its
topology in between two periodic messages i.e. in between Tpr and Tpr +1 say at T time. The
routing protocol will not wait for next periodic message to update this change in topology
instead; it immediately broad cast a triggered message. The following illustration will help in
clearing the concept of triggered message.
50
Analytically we can express this illustration as
Tpr<T<Tpr+1
As discussed in [34] this notation can be expressed as:
Tr(a) = βT\Tprβπ/πππ
β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦15
Where β β in mathematics, is the ceiling operator and it should be solved by taking the highest
possible values. In a network where only one node a moves within a time span of Tpr and
Tpr+1 , the above equation qualifies but considering a highly mobile environment where all
the nodes of network are mobile, the maximum over head due to triggered update during Tpr
and Tpr+1 will be:
OH(Tr) = β βT\Tprβππππ
ni=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦..16
51
Where
OH(TR) = Trigger message over head
T = triggered update
3.2.4 Routing Overhead of Proactive Routing Protocol:
Placing the respective values of Eq.2, Eq.3 and Eq. 5 in Eq. 1 we get the analytical equation
expressing the total routing overhead of the network.
π π = οΏ½ οΏ½ οΏ½Qrl (Tpr)Na(TPR)οΏ½ + kn3
π΅βπππ
L1
π=0ππβππ΄
+ β βT\Tprβππππ
ni=1 β¦β¦β¦β¦β¦β¦β¦β¦17
Let RO be the optimized function y (Β΅k, Tpr, T, n, Ξ»).Hence Eq.6 can be written as:
π¦(Β΅k, Tpr, T, n, Ξ») = οΏ½ οΏ½ οΏ½Qrl (Tpr) Ξ»(Tpr)οΏ½ + kn3
π΅βπππ
L1
π=0ππβππ΄
+ β βT\Tprβππππ
ni=1 β¦β¦..β¦.18
As discussed in [35]:
Qrl (Tpr)ππ₯ = 1 β πβππππ/Β΅πβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦.19
And we can say that Ξ» is the average number of packets arrived successfully at node, Β΅k is the
uplink time, T is triggered update messages and n stands for number of nodes in a network.
Substituting the value from Eq.8 to Eq.7, we get our Eq. 9
π¦(Β΅k, Tpr, T, n, Ξ») = (PNavg)οΏ½ οΏ½(1 β πβππππ/Β΅π) Ξ»(Tpr)οΏ½ + kn3
π΅βπππ
L1
π=0+ β βT / Tprβ
ππππ
ni=1 ..20
52
To analyze the variation in these parameters, we take the partial derivative of function y and
we get:
πΏπ¦πΏπππ
= (C)οΏ½ οΏ½οΏ½1 β πβππππ
Β΅π οΏ½ + οΏ½ rβTprΒ΅k
οΏ½ πβππππ
Β΅π οΏ½ β kn3
π΅βTpr2
Lavg
π=0
+ β οΏ½βT\Tpr2οΏ½+T\Tpr2
π2/πππ2ni=1 β¦21
Take PNavg* Ξ»= C the partial derivative with respect to the rate of packet arrival will be
expressed as eq 11
πΏπ¦πΏπ
= (PNavg)οΏ½ οΏ½Tpr οΏ½1 β πβππππ
Β΅π οΏ½οΏ½Lavg
π=0
β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦22
Similarly if we take partial derivative with respect to T, we will get the eq 12.
πΏπ¦πΏπ
= β οΏ½1\Tpr2οΏ½β1\Tpr2
π2/πππ2ni=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦23
If we want to analyze the rate of change of link arrival time in the same scenario we will get
eq 13.
πΏπ¦πΏΒ΅π
= (PNavg β Ξ» β Tpr)οΏ½ οΏ½βrTprΒ΅k2
οΏ½πβππππ
Β΅π οΏ½οΏ½Lavg
π=0
β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦24
and obviously number of nodes of a network plays a vital role in routing over head. We can
calculate the impact of change in number of nodes of a network we can do so via eq 14.
πΏπ¦πΏπ
= 3ππ2
π΅βπππβ¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦...25
53
3.2.5 Calculations and Results:
Considering the Eq.14, that is partial derivative w.r.t. β nβ, we can infer that as number of
nodes of a network increases, its routing overhead increases but, if number of nodes decreases
than 3 nodes, the over head of the network will reduce.
Keeping in view Eq. 10 and Eq. 12 are the partial derivatives with respect to Tpr and T i.e.
periodic update interval and triggered updates. Considering all the aspects of network in
mobility and scalability remain constant, than we can say that these two variables are
dependent on each other. As the periodic message exceeds in its interval, there would be more
triggered messages than before. In the same sense if we reduce the periodic update interval
time, triggered updates will be lowered but only and only if all other parameters mainly
mobility and number of nodes in a network remain same. To further analyze this change, we
take total derivative of Tpr and T variables of function y.
ππ¦ππππ
=πΏπ¦πΏπππ
+πΏπ¦πΏπ
οΏ½dT
dTprοΏ½
ππ¦ =πΏπ¦πΏπππ
(ππππ) +πΏπ¦πΏπ
(dT)
ππ¦ =
(C)οΏ½ οΏ½οΏ½1 β πβππππ
Β΅π οΏ½ + οΏ½ rβTprΒ΅k
οΏ½ πβππππ
Β΅π οΏ½ β kn3
π΅βTpr2
Lavg
π=0
+ β οΏ½βT\Tpr2οΏ½+T\Tpr2
π2
πππ2
ni=1 (ππππ) +
β οΏ½1\Tpr2οΏ½β1\Tpr2
π2
πππ2
ni=1 (dT)β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦26
54
In static environment, longer Tpr might not affect the performance of routing protocol and
favor reducing routing over head but if in mobile environments, if longer Tpr is used, it will
result in high rate of triggered messages. Eq. 15 shows that Tpr and T are two variables which
are tied with nonlinear relationship with one another. Simulation results will clearly validate
this mathematical statement.
Considering Eq.11 and Eq. 13, it is obvious that if there is always an uplink for whole of the
network / for the entire life of the network, or there is no periodic interval i.e ifΒ΅k tends to
infinity and Tpr is zero, both partial derivatives with respect to Ξ» and Β΅k will be zero. [33]
Assuming πΏy / πΏTpr = 0, we get:
(C)οΏ½ οΏ½οΏ½1 β πβππππ
Β΅π οΏ½ + οΏ½ rβTprΒ΅k
οΏ½ πβππππ
Β΅π οΏ½Lavg
π=0
= kn3
π΅βTpr2β β οΏ½βT\Tpr2οΏ½+T\Tpr2
π2/πππ2ni=1 β¦β¦β¦27
The ratio between periodic update time and uplink time can be termed as update coefficient
[33]. Let us denote that update coefficient as βhβ= Tpr/ Β΅k or Tpr= Β΅k*h.
Placing the values in Eq. 14 will give us optimized network analytical model.
(C)οΏ½ {(1 β πβπβ) + (r β h)πβπβ}Lavgπ=0 = kn3
π΅β(Β΅πββ)2β β οΏ½βT\(Β΅πββ)2οΏ½+T(Β΅πββ)2
π2/(Β΅πββ)2ni=1 β¦β¦β¦β¦.28
Eq. 17 shows that if average link uptime increases, βhβ or update coefficient will also increase
but this increase donβt linearly affect the periodic interval time. As depicted before, here again
this equation shows the same, as number of nodes increases, the routing overhead will also
nonlinearly increase.
This is the generalized routing over head analytical model for proactive routing protocols. In
this model we have taken care of DSDV by injecting trigger message which were not
discussed in [33] and [35], where as in many other papers including [34], periodic messages
55
and triggered messages are taken into consideration leaving the overhead due to failure of
packets in reaching at destination. In our model, we consider all these three vital factors
influencing routing overhead.
If we consider OLSR routing protocol, than we will come to know that it is actually an
optimization of LSR with the use of MPR concept. MPR set is a subset of set of all the nodes
of network. As small the MPR set will be, as much over head or retransmissions will be
reduced.
According to RFC 3626 [36], there are 4 periodic messages in OLSR. Hello, Topology
control, MID and HNA. Mostly only hello message and TC messages are taken into
considerations. If we look into theme of OLSR routing protocol, we will come to know that
hello messages are used to gain the neighborhood knowledge and to select MPR set (MPR set
is the only set which is allowed to retransmit or broadcast the receiving message). once,
neighborhood is known and MPR set is established, TC message is broadcasted by that
selected MPR set for topology information. Hence Hello message as given in RFC3626, is of
1 sec while TC message interval is 2 sec. in other words, authors of RFC 3626 proposed that
hello message interval should be taken as half of the TC message interval.
Placing the values of hello message and TC message in composed analytical model, we get the
equation:
π πππΏππ = (PNavg)οΏ½ οΏ½(1 β πβππππ/Β΅π) Ξ»(Tpr)οΏ½+ kn3
π΅βπ»
L1
π=0+ kn3
π΅β2π»+ β βT/H+2Hβ
ππ»+2π»
ni=1 β¦β¦..29
Where
H = Hello message interval
56
2H = TC message interval, twice the
To analyze the rate of change in hello and TC interval, we partially derivate equation 18 with
respect to βhβ and we get:
πΏπ¦πΏπ»
= β ππ3
π΅βπ»2 βππ3
π΅β2π»2 + β οΏ½βT/(H+2H)2οΏ½+T\(H+2H)2
π2/(H+2H)2ni=1 β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦β¦30
With the help of same model, we can also calculate our desired over head, whether to find
over head due to hello message emissions or topology control messaging over head or the
overhead due to lost packets.
3.3 Future works:
`Considering the step 4 of initially described steps in RD and RM phases of reactive routing, a
reactive routing protocol may choose the following three options [28] i.e.
1. Wait for TIME OUT time and than new route discovery
2. RERR packet to main source node via reverse path and new route discovery
3. Local repair
We will find the overhead considering all the three phases one by one. Especially the case 3 is
the requirement of the time.
57
Chapter 4
Experiments, Results and Discussions
58
4.1 Simulations
AODV [38], DSR [40], DSDV [39] are used as default ones that come with NS2. AODV
coding used was developed by CMU/MONARCH group while it was optimized by Samir Das
and Mahesh Marina (University of Cincinnati). DYMOUM by MASIMUM [41] is used for
DYMO. We used NS-2.34 for simulating AODV and DSR while, DYMOUM was simulated
in NS 2.29. Sven JAAP [42] gives us FSR implementation while MASIMUM produced UM-
OLSR [41] which we utilized here. Simulations of DSDV and OLSR are made on NS-2.34
while FSR was already available on NS-2.30. We pin points on mobility and Scalability
factors of Ad Hoc networks in this work. Hence all the simulations are done taking these
aspects into account.
4.2 Parameters for Experiments:
We have considered a network of 50 nodes where nodes are randomly located and mobile.
These nodes are privileged with the bandwidth of 2 Mbps. Speed by which a data packet
travels is set at 8m/s. Packet size is defined as 512 bytes while simulation setup runs on
Continues bit rate traffic sources. The size of network is defined as 1000 square meter.
For every experiment, any particular scenario with same parameters and pause timings is
executed for 5 times. The mean value of these 5 times is used to plot the graphs which are
presented underneath.
βcbrgenβ located in directory/ nsallinone-nn/ns-nn/indep-utils/cmu-scen-gen is used in making
the random source to destined node interconnections . The traffic we used under such scenario
is CBR i.e. constant bit rate.
Given these parameters, we have confined our experiments only for the three metrics.
59
1. Throughput
2. End - End Delay
3. Normalized routing Load.
4.3 Reactive Experimentations:
4.3.1 Through put of Reactive Routing:
a) With respect to mobility
In general sense, throughput refers to the amount of data that has successfully reached its
destination. Mathematically it can be stated as
πβπππ’πππ’π‘ =(ππ’ππππ π ππ’πππ¦ πππππ£ππππ πππ‘π ππππππ‘π )
ππππ
Represented as B/s
Considering simulation results for throughput, we can state that when mobility factor in a
network is high, DSR enjoys the maximum throughput with respect to AODV and DYMO
subjected to FCM not at 0, 100, 200 pause times. If FCM is at 0 or 100 or 200 pause times,
AODV gives better throughput than DYMO and DSR. If mobility factor is further increased,
DSR will face trouble in route caching (as no mechanism is defined to erase stale routes hence
it fails to converge at high mobility). If we consider AODV, than it surely have a TIME_OUT
factor involved. AODV waits for a specified time and than the route is termed invalid finally
erased from routing table. βHello messagesβ in AODV also works very well for high mobile
environment. Leaving highly mobility factor, DSR performs most stable throughput as no
unnecessary packets are generated by this routing protocol. In DSR, in link breakages, it also
have multiple routes for the same source and destined station while, in AODV, the routing
table keeps only one the best chosen path between the nodes. Hence the environment where
links are immune to break due to noise or other factors, DSR again supersedes AODV and
DYMO. DYMO proves itself the worst amongst the other two protocols.
60
Throu
ghpu
t (Kbp
s)
Pause Time
Fig. 4.1 Throughput (AODV, DSR, DYMO)
Throu
ghpu
t (Kbp
s)
No. of Nodes
b) With respect to scalability:
According to our experiments, AODV converged almost all data rates with all
scalabilities, While DSR proves itself to be scalable but during high data traffic, it
couldnβt converge the network. DYMO perfoms worst among these studied routing
protocols. As the number of nodes increases or data traffic increases, its performance
degrades dramatically. According to [18] a network of multiple thousands of nodes
with different traffic loads can be handled by AODV. The reason that AODV
supersedes DSR and DYMO is lower packet loss ratio and propagating of information
regarding distant vector which practically consume minimum bandwidth. This feature
gives AODV the room for scalability and traffic loads across the number of routes. In
AODV, routing packet contains only one hop information while in DSR; packet size is
61
larger as it keeps the information of whole route. This is another reason that AODV
outperforms DSR.
4.3.2 End - End Delay of Reactive Routing:
a) With respect to mobility
The time a packet takes in reaching destined node from the originator node can be termed as
End - End delay. Mathematically we can express End - End delay as
πΈπ· =ππ’ππππ ππ ππππππ‘π π‘ππππ πππ‘π‘ππ(π ππ’ππ π‘πππ π‘πππ)
ππ’ππππ ππ π π’ππππ π ππ’πππ¦ πππππ£πππ ππππππ‘π
End - End delay is calculated in time scale i.e. seconds.
As shown in the graph, AODV produces lowest performance as link breakages due to mobility
factor may lead to longer routes. DYMO though works worst in throughput case but here it
works best amongst DSR and AODV. It is so because, DYMO donβt check the routes in
memory as DSR looks into route cache and AODV in to its routing table, instead DYMO
starts expanding ring search algorithm whenever a route is required.
b) With respect to scalability:
The concept of grat. RREP is used both in DSR and AODV while this is the reason that
DYMO enjoyes lowest End - End delay irrelevant of number of nodes in the network. Grat.
Route replies though results in lower delay at normal traffic rates but DSR checks the route
cache before using expanding ring search algorithm in the same way as AODV searches the
route in its routing table before starting a route request using expanding ring search algorithm.
DYMO donβt use such stored information rather it simply initiates ERS. AODV also have a
link repair feature that makes it bear the highest End - End delay with respect to any
scalability among DYMO and DSR.
62
No. Of Nodes
End t
o End
Delay
(ms)
Pause Time
Fig. 4.2 End to End Delay (AODV, DSR, DYMO)
End t
o End
Delay
(ms)
4.3.3 Normalized Routing Load of Reactive Routing:
a) With Respect to Mobility:
When a single data packet is to be sent from one node to another within an ad hoc network, a
number of routing packets are involved in sending this data packet . The number of these
routing packets which are sent just to transfer one data packet are termed as normalized
routing load. Mathematically we can state it as
ππππππππ§ππ π ππ’π‘πππ πΏππ΄π· = πππ‘ππ π ππ’π‘πππ πΏπππ β ππ’ππππ ππ πππ‘π ππππππ‘π π πππ‘
AODV and DSR use the concept of gratuitous route replies. i.e. when a route request reaches
any node that has a valid route stored in its route cache or routing table, it generates a route
reply by its self to the main source node. This route reply contains the full information upto
the destined node and overhead of finding route beyond that node limits. DYMO donβt use
this grat. RREP. And thatβs the reason; DYMO suffers from greater routing overhead with
respect to the other two protocols. AODV also works well in the context of normalized routing
overhead but, there is a concept of local link repair and above all the use of hello message for
63
link monitoring makes it perform lowers then DSR. A node with underlying DSR protocol use
promiscuous mode and this is the reason that it bears lowest over head.
A common observation with respect to increase in mobility of nodes in network, all the three
routing protocols bear gradually higher overhead. The reason is propagations of route error
packets. As the mobility increases, the chances of link breakages also increase in the same
proportion. Hence routing overhead must be increased.
No. of Nodes
Norm
alized
Routin
g Loa
d
Pause Time
Fig 4.3 Normalized Routing Loead (AODV, DSR, DYMO)
Norm
alized
Routin
g Loa
d
b) With respect to scalability:
In normal conditions of network populations or even at high population, routing overhead of
DYMO is lower than that of AODV and DSR where as AODV bears high routing overhead in
dense networks. Periodic link sensing, packets involved in local link repair mechanism and
grat. RREP results in high routing overhead. Whereas promiscuous mode utilized by DSR
reduces the routing overhead in no so dense environment.
.
64
4.4 Proactive Experimentations:
4.4.1 Throughput of Proactive Routing:
a) With respect to mobility:
When we simulated proactive routing protocols for throughput prospective, DSDV proves
itself to be the most efficient protocol among other two i.e. FSR and OLSR. Main reason of
this result is the basic functioning of DSDV protocol. As stated in chapter 3 as well that a
packet is sent only on the best possible route and this route search may get a bit of delay as
DSDV delays the route advertisements that have higher probability to be changed earlier.
Moreover, un-stabilized routes that have the same sequence number in DSDV routing protocol
are also advertised with delay. These features of DSDV results in accurate routing hence
increased throughput. On other hand, taking OLSR into account, its ability to converge
declines as the mobility increases, thus lower throughput is the result. Though, in static
environment, due to MPR concept in OLSR, it gives better throughput than FSR and DSDV.
Whenever a link breaks, there is a concept of triggered messages in DSDV routing protocol
that also increase the route accuracy where as in FSR and OLSR there is no availability of
triggered updates.
b) With respect to scalability:
FSR is better under fluctuating traffic patterns while OLSR works well with dense
environment among proactive routing protocols. Throughput of FSR also increases as it used
multi level fisheye scope. This technique results in lower overhead, and less consumption of
bandwidth which is a major plus point for the throughput. DSDV use NPDUs for lower
overhead but triggered messages create routing overhead consuming bandwidth and resulting
in lower throughput. OLSR uses MPR for lowering the routing overhead but periodic
messages used to calculate and compute a MPR set for a node take more bandwidth. Though
65
its throughput is more than that of DSDV but lower than FSR. FSR, is highly scalable as it use
different frequencies for different scopes i.e. at different time intervals. OLSR is also highly
scalable but network convergence is issue in very dense environment where traffic load is
high. Hence OLSR is a highly scalable protocol but if data traffic rates are not that high. Th
rough
put
No. of Nodes
Pause time
Fig 4.4 Troughput (FSR, DSDV, OLSR)
Throu
ghpu
t
4.4.2 End - End Delay of Proactive Routing:
a) With respect to mobility:
According to the general behavior of proactive routing it can be observed that End - End delay
will be higher with respect to reactive routing protocols when we emphasis on mobility of
nodes within the network. The reason is spreading the topology and routing information to
each and every participant node of the network. Whenever any change occurs, it is notified to
entire network. DSDV proved to be the best for throughput but when considering delay, it
bears the worst conditions with respect to FSR and OLSR. DSDV, donβt sends the data packet
66
until a fair enough time for creating and choosing the best possible route. Moreover the
delayed advertisements of unstable routes results in overall End - End delay. In DSDV, this is
done to reduce the routing overhead and provide route accuracy but it compromises on delay.
OLSR performs better than DSDV. FSR owns the highest End - End delay among the studied
protocols. As in the basic theme of FSR, when the mobility increases, the accuracy of far away
destined nodes fades. However as the packet gets closer to destined node, the routing info gets
accurate.
No. Of Nodes
Pause Time
End t
o End
Dela
y (ms
)
Fig 4.5 End to End Delay (DSDV, FSR, OLSR)
End t
o End
Dela
y (ms
)
b) With respect to scalability:
As the environment get dense, the End - End delay of all three routing protocols i.e. FSR,
DSDV and OLSR will increase. FSR shares routing updates with its neighbors in small
intervals while the informationβs shared at far away nodes has some larger interval in them. As
the network become more scalable, End - End delay will increase in FSR. In DSDV, End -
End delay is due to the two procedures, i.e. finding some routes, choosing the best one route.
As the network gets denser, the routing of End - End time will also increase. As in proactive
67
nature, the information is spread in whole network. OLSR use MPRs and during low traffic
rates, End - End delay using OLSR will be lower. This is because of MPR concept that
presents weel organized flooding control instead of flooding the packet on whole network.
4.4.3 Normalized Routing Load of Proactive Routing:
Norm
alized
Routin
g Load
No. Of Nodes
Pause Time
Fig 4.6 Normalized Routing Load (FSF, DSDV, OLSR)
Norm
alized
Routin
g Load
a) With Respect to Mobility:
Among the studied proactive routing protocols, OLSR bears maximum number of periodic
messages to compute MPRs. Hence as it was observed, experiments proved it as well that
OLSR bears highest routing overhead. DSDV again proves to be the normal choice amongst
FSR and OLSR in terms of routing overhead. Reason is discussed in End - End delay section
of proactive routing protocols. Considering FSR, it bears lower overhead due to control and
periodic messages in competition with OLSR. FSRβs control messages are periodic based
rather event driven based as in OLSR. This feature helps FSR in reduced routing overhead.
68
Moreover, there is limited flooding in FSR i.e. link state information is not flooded among
whole network besides, every node manage a LS table which is derieved on the basis of UP
TO DATE information received. This information is not broadcasted or flooded but is shared
amongst neighbors.
b) With respect to scalability:
Among all studied proactive routing protocols, OLSR gives the highest routing overhead due
to MPR comupational messages and topology control messages. DSDV and FSR have lower
overhead in dense environments. DSDV reduces overhead with the help of NPDUs and FSR
use Fisheye scope to limit overhead. The simulated results show that FSR stands best amongs
DSDV and OLSR in a dense and highly mobile environment in terms of overhead.
4.5 Conclusions:
In this thesis, we have initially presented the literature survey of route discovery and
maintenance phases of both reactive and proactive routing protocols for mobile Ad Hoc and
Vehicular Ad Hoc networks. We contributed generalized analytical framework representing
routing overhead due to route discovery and maintenance phases of both reactive and
proactive routing protocols. Reactive routing is based on immediate response. This can waste
network resources. Binary exponential back off algorithm and Expanding ring search
algorithms are used in AODV DSR and DYMO. Both of these algorithms are expressed in
this thesis as well. Later experiments are conducted on the six routing protocols keeping the
parameters of throughput, End - End delay and routing overhead in emphasis with respect to
mobility and scalability factors of any Ad Hoc network. Routing overhead comprises of the
overhead caused by control packets only. Data packets are not included in routing overhead.
As the routing overhead increases and or End - End delay increases, performance of the
routing protocol deteriorates. The protocol that uses minimum resources of bandwidth by its
69
control packets can provide better data flow. Hence the environments where traffic load is
very high, only those protocols that have low routing overhead will survive. If we consider
scalability, than AODV stands at top of rest of five routing protocols. It use distance vector
distribution that minimizes the network resources consumption. In proactive routing, OLSR
stands tall as it limits retransmissions due to use of MPR concept but only in dense
environments. If mobility with the number of nodes of network increases, than FSR is be a
good choice as it generates low routing overhead that leads to high data rates within the
limited bandwidth.
The experiments conducted show us that considering over all behaviors and performance of
routing protocols, reactive protocols works better than proactive routing protocols if only
mobility factor is considered. The network underlying AODV protocol, bears low routing
overhead as control packets of AODV contains a very small part of information in them
where as if we compare it with DSR, control packet of DSR carries whole routing
information in it. Hence we can say that DSR has higher routing overhead in terms of bytes
or size. If we consider number of control packets than DSR broadcast lesser number of
packets than that of AODV. AODV use periodic hello packet for link sensing and also the
local repair routing overhead. Hence if we genially compare both of these routing protocols
considering mobility and speed factors, we can conclude that both of these protocols give
almost same performance.
Concluding all the six routing protocols, our study suggest that, AODV can be selected for
denser environments where lower routing overhead is required, DSR should be used within a
network having limited number of hops. DYMO routing protocol can be used in networks
where delay is in tolerable. As like other reactive protocols, DYMO donβt look for any stored
route as DSR looks into its cache and AODV in its routing table. It initializes binary
70
exponential back off and expanding ring search algorithm immediately. In proactive routing,
OLSR also stands good for lower end to end delay due to MPRs. The following table
summarizes the basic differences and similarities between the studied six routing protocols.
AODV DSR DYMO FSR OLSR DSDV
Protocol type Distance Vector
Source routing
Source routing
Link state Link state Distance Vector
Route maintained in
Routing table Route Cache Routing table Routing table Routing table Routing table
Multiple route discovery
No yes no yes no No
Multicast Yes yes no yes yes Yes
Periodic broad cast
Yes no no Yes (with in limited range)
yes Yes
Route reconfiguration
Erase route notify source
Erase route notify source
Erase route notify source
Link state mechanism with sequence number
Link state mechanism / Control messages send in advance
Sequence number adopted
Route discovery packets
using RREQ and RREP packets
using RREQ and RREP packets
using RREQ and RREP packets
Link state messages
via control message link sensing
Via control messages
Limiting over head ,collision avoidance, network congestion
Expanding ring search algorithm using specific time period
Expanding ring search algorithm using hop count
Expanding ring search algorithm using specific time period
Fisheye procedure, broadcast limited only to transmission range.
Concept of Multipoint relays
concept of sequence number
Limiting over head , collision avoidance , network congestion
Binary exponential back off time
Binary exponential back off time
Binary exponential back off time
MAC layer protocols only
MAC layer protocols only
MAC layer protocol only
71
Update information
By RERR message
By RERR message
By RERR message
Only neighbor information
2 hop neighbor information
By control messages
Topology information
Full topology Full topology Full topology Reduced topology
Full topology Full topology
Update destination
Source source source neighbors MPR sets source
Broadcast Full Full Full Local/ limited Limited by MPR Set
full
Reuse of routing info
No Yes Yes Yes Yes yes
Route selection Only searched route
Hop count Only searched route
Shortest hop count
Hop count Shortest hop count
72
References:
[1] Xiaogeng Zhao, ββAn Adaptive Approach for Optimized Opportunistic Routing Over Delay Tolerant Mobile Ad Hoc Networksββ, Computer Science Department, December 2007
[2] Ph.D. Thesis of Nadeem Javaid, βAnalysis and Design of Link Metrics for Quality Routing in Wireless Multi-hop Networksβ, University of Paris-Est, 2010.
[3] S. Chen and K. Nahrstedt, βDistributed quality-of-service routing in Ad-hoc networks,β IEEE J. Select. Areas Commun., vol. 17, no. 8, pp. 1488β1505, 1999.
[4] C. E. Perkins and E. M. Royer, βAd-hoc on-demand distance vector routing,β in Proc. 2nd IEEE Workshop on Mobile computing Systems and Applications (WMCSA β99), pp. 90β100, New Orleans, La, USA, February 1999.
[5] D. B. Johnson and D. A. Maltz, βDynamic source routing in Ad-hoc wireless networks,β in Mobile Computing, T. Imielinski and H. Korth, Eds., chapter 5, pp. 153β181, Kluwer Academic, Norwell, Mass, USA, 1996.
[6] C. E. Perkins and P. Bhagwat, βHighly Dynamic Destination- Sequenced Distance-Vector Routing (DSDV) for Mobile Computers,β SIGCOMM, London, UK, August 1994, pp. 234-244.
[7] D. B. Johnson and D. A. Maltz, β Dynamic Source Routing in Ad-Hoc Ad hoc Networks," Mobile Computing, ed. T. Imielinski and H. Korth, Kluwer Academic Publishers, 1996, pp. 153-181.
[8] C. Perkins, Ad hoc On demand Distance Vector (AODV) routing, IETF Internet draft (1997), http://www.ietf.org/internet-drafts/draftietf- manet-aodv-00.txt.
[9] Antipapal Suresh. Performance Analysis of Ad hoc On-demand Distance Vector routing (AODV) using OPNET Simulator [J].Communication Networks University of Bremen, 2005, 10οΌ5οΌ:63-69.
[10]I. Chakeres, C. Perkins. βDynamic MANET On-demand (DYMO) Routing draft-ietf-manet-dymo-21,July 2010.
[11]I. Chakeres, CenGen, WiChorus, etc. Dynamic MANET On-demand (DYMO) Routing[Z]. Intemet-Draft, draft-ietf-manet-dymo-19 (work in progress), March 22, 2 010.
[12]Royer E M, Toh C K. A review of current routing protocols for Ad Hoc mobile wireless networks [J]. IEEE Personal Communications, 1999 6(2): 46- 55.
[13] Vincent D. Park and M.Scott Corson. A highly adaptive distributed routing algorithm for mobile wireless networks. In Proceedings of INFOCOM 1997, 1997.
73
[14] D. B. Johnson and D. A. Maltz, βRFC 4728, The Dynamic Source Routing Protocol (DSR) for Mobile Ad Hoc Networks for IPv4β feb 2007
[15] John Jubin and Janet D. Tornow. The DARPA Packet Radio Network Protocols. Proceedings of the IEEE,75(1):21-32, January 1987
[16] Ph.D. Thesis of Nadeem Javaid, βAnalysis and Design of Link Metrics for Quality Routing in Wireless Multi-hop Networksβ, University of Paris-Est, 2010.
[17] G. Pei, M. Gerla, and T.-W. Chen, "Fisheye State Routing: A Routing Scheme for Ad Hoc Wireless Networks", Proceedings of ICC 2000, New Orleans, LA, Jun. 2000.
[18] L. Kleinrock and K. Stevens, "Fisheye: A Lenslike Computer Display Transformation," Technical report, UCLA, Computer Science Department, 1971.
[19] Mario Gerla, Guangyu Pei, Xiaoyan Hong, and Tsu-Wei Chen, \Fisheye state routing protocol (FSR) for ad hoc networks," Internet Draft http://www.ietf.org/internet-drafts/draft-ietf-manet-fsr-01.txt, November 2000, Work in progress
[20] Atsushi Iwata, Ching-Chuan Chiang, Guangyu Pei, Mario Gerla, and Tsu wei Chen, \Scalable routing strategies for ad hoc wireless networks," Tech. Rep., Department of Computer Science University of California, Los Angeles, 1999.
[21] L.R. Ford Jr. and D. R. Fulkerson. Flows in Networks. Princeton Univ. Press, 1962.
[22]. Guoyou He ,βDestination - Sequence Distance Vector (DSDV)β Networking Laboratory. Helsinki University of Technology. 1999.
[23] Qayyum, A.; Viennot, L.; Laouiti, A.; , "Multipoint relaying for flooding broadcast messages in mobile wireless networks," System Sciences, 2002. HICSS. Proceedings of the 35th Annual Hawaii International Conference on , vol., no. pp. 3866- 3875, 7-10 Jan. 2002 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=994521&isnumber=21442
[24] P. Jacquet, P. Minet, P. Muhlethaler, N. Rivierre. Increasing reliability in cable free radio LANs: Low level forwarding in HIPERLAN. Wireless Personal Communications, 1996.
[25] A. Qayyum, L. Viennot, A. Laouiti. Multipoint relaying: An efficient technique for flooding in mobile wireless networks. 35th Annual Hawaii International Conference on System Sciences (HICSSβ2001).
[26]Zheng Li; Nenghai Yu; Zili Deng; , "NFA: A New Algorithm to Select MPRs in OLSR," Wireless Communications, Networking and Mobile Computing, 2008. WiCOM '08. 4th International Conference on , vol., no., pp.1-6, 12-14 Oct. 2008 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4678560&isnumber=4677909
74
[27] Mohammad Naserian; Kemal E. Tepe; Mohammed Tarique; βRouting Overhead Analysis for Reactive Routing Protocols in Wireless Ad Hoc Networksβ, Wireless And Mobile Computing, Networking And Communications, 2005. (WiMob'2005), IEEE International Conference on , vol.3, no., pp. 87- 92 Vol. 3, 22-24 Aug. 2005
[28] Ionut d. Aron ; Sandeep K.S. Gupta; βAnalytical Comparison of Local and End-to-End Error Recovery in Reactive Routing Protocols for Mobile Ad Hoc Networks β Proc. ACM Workshop Modeling, Analysis, and Simulation of Wireless and Mobile Systems (MSWiM), 2000.
[29] C. Perkins, E. Belding-Royer, and S. Das. Ad-hoc on-demand distance vector (AODV) routing, Internet RFC 3561, July 2003.
URL: http://tools.ietf.org/html/rfc3561
[30]R. Negi and A. Rajeswaran, Capacity of power constrained ad-hoc networks, Proceedings of IEEE Infocom, 2004.
[31]S. Basagni, M. Conti, S. Giordano, I. Stojmenovic, Ad Hoc Networks Technologies and Protocols, IEEE Press, Weliy Interscience, 2004.
URL: http://www.citeulike.org/user/HailunTan/article/106736
[32] Nadeem Javaid et al.β Modeling routing overhead Generated by Wireless Proactive Routing Protocols,β USA Globecom 2011.
URL: http://arxiv.org/PS_cache/arxiv/pdf/1108/1108.3703v1.pdf
[33]Pan-long Yang; Chang Tian; Yong Yu; , "Analysis on optimizing model for proactive ad hoc routing protocol," Military Communications conference, 2005. MILCOM 2005. IEEE , vol., no., pp.2960-2966 Vol. 5, 17-20 Oct. 2005 doi:10.1109/MILCOM.2005.1606113F URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1606113&isnumber=33743
[34]Xianren Wu; Sadjadpour, H.R.; Garcia-Luna-Aceves, J.J.; , "Routing Overhead as A Function of Node Mobility: Modeling Framework and Implications on Proactive Routing," Mobile Adhoc and Sensor Systems, 2007. MASS 2007. IEEE Internatonal Conference on , vol., no., pp.1-9, 8-11 Oct. 2007 doi: 10.1109/MOBHOC.2007.4428595 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4428595&isnumber=4428592
[35] Ben Liang and Zygmunt J. Hass, Optimizing Route-Cache Lifetime in Ad Hoc Networks, Proceedings of the 22nd IEEE INFOCOM April 1st 3rd ,2003.
[36] T. Clausen and P. Jacquet, .RFC 3626: Optimized link state routing protocol OLSR),October 2003.URL: http://tools.ietf.org/html/rfc3626
75
[37] Mario Gerla, UCLA Guangyu Pei Xiaoyan Hong, UCLA Tsu-Wei Chenβ Fisheye State Routing Protocol (FSR) for Ad Hoc Networksβ βdraft-ietf-manet-fsr-01.txtβ, November 17, 2000
[38] C. M. University, βAODV Copyright (c) 1997, 1998 Carnegie Mellon University, http://www.cmu.edu/,β 1998. [39] CMU/Monarch, βDSDV Copyright (c) 1997 Regents of the University of California, http://www.universityofcalifornia.edu/regents/ β 1997. [40] CMU/Monarch, βDSR Copyright (C) 2000 by the University of Southern California. http://www.usc.edu/,β 2000 [41] University of Murcia,βMASIMUM (MANET Simulation and Implementation at the
University of Murcia)β 2009.
[42] The Institute of Operating Systems and Computer Networks,βFSRβ. http://www.ibr.cs.tu-bs.de/ 2009.
76