Qureshi_Umair_Saeed.pdf (3.7 MB)
Download file

A Performance Study of TCP on Ad Hoc Networks

Download (3.7 MB)
thesis
posted on 08.06.2021, 07:56 authored by Umair Saeed Qureshi

Ad-hoc networks, characterized by highly dynamic multi hop wireless cormectivity, offer challenges related to unique issues of congestion, channel error, routing instability and network partitioning.  Dealing with these issues requires precise detection of network states, which we accomplished by measuring appropriate metrics, such as packet out of order,  inter-arrival  delay  differences,  connection throughput,  round  trip  time  etc.  We evaluated the performance  of TCP  under variety  of network  conditions  running  two important  routing  protocols  namely  Dynamic  Source  Routing  (DSR)  and  Dynamic Sequential Distance Vector (DSDV) routing. These protocols belong to different class of routing protocols. DSR is an on-demand whereas DSDV is a link-state routing protocol. In this project, we carried out detailed simulations of a sizable ad-hoc network using NS2 to study the dynamics of the two routing protocols related to the performance of TCP by calculating the above metrics. We observed that congestion in ad-hoc network exhibits dynamic behavior and sometime it is not as bad as in case of fixed networks. For example we  observed  that  node  mobility  introduces  transience  to  congestion  by  dissipating congestion at bottleneck nodes. We observed in at least one scenario that node movement totally avoids congestion. We evaluated the performance under channel error conditions by measuring packets out of order and packet losses for both protocols. We also studied the routing characteristics of both protocols under identical mobility conditions. Finally, we evaluated the worst-case performance under extreme network condition by combining congestion, channel error and node mobility. 

History

Language

eng

Degree

Master of Engineering

Program

Electrical and Computer Engineering

Granting Institution

Ryerson University

LAC Thesis Type

Thesis Project

Thesis Advisor

Dr. Jaseem-ud-Din Xiao-Ping Zhang