Congestion control enhancement over wireless networks

Abstract

In this thesis, we analyze the performance of wireless LAN networks subject to random loss. In this regard, we use a congestion control technique that has been introduced in a previous study, titled TCP Congestion Control Enhancement for Random Loss (TCP CERL). TCP CERL is a sender-side modification of TCP Reno protocol. TCP CERL is an end-to-end technique that achieves high performance over wireless links and does not decrease the congestion window and slow start threshold if the random loss is detected. TCP CERL assumes a static threshold (A) equal to 0.55 which does not perform well when considering a heavy traffic load compared to new protocols. In this thesis, we propose a modified version of TCP CERL, called TCP CERL PLUS (TCP CERL+, in short). TCP CERL+ works similarly as TCP CERL, but its main idea is to use a dynamic threshold (A) in terms of Round-Trip Time (RTT) rather than static threshold. By doing so, we employ the average RTT and the minimum RTT measurements made over the connection to evaluate the queue length of the bottleneck link. In this thesis, we compare TCP CERL and TCP CERL+ with TCP New Jersey+, TCP mVeno, TCP Westwood+, TCP Cubic, TCP YeAH, and TCP NewReno by using Network Simulation NS-2. The results will show that CERL+ outperforms CERL when there are many users and Two-way transmission.