A Throughput Analysis of TCP Variants in Mobile Wireless Networks

Abstract

Unlike wired networks, random packet loss due to bit errors is not negligible in wireless networks. These networks have mobility as another challenge. This causes significant performance degradation of TCP. Different Variants of TCP exist; some have no mechanism to differentiate random losses on the wireless link from congestion and therefore treat all random losses as an indication of congestion. Analyzing the performance of existing TCP Variants under random packet loss is important in order to select an appropriate version of TCP in the networks where random packet losses are more frequent. This paper describes an ns-2 based simulation analysis of TCP-Vegas, TCP-Tahoe, TCP-Reno, TCP-New Reno, TCP-SACK, TCP-Veno, TCP-Westwood, TCP-Westwood New Reno and TCP-New Jersey under random packet loss rate in wireless networks. Experimental studies indicate that the Throughput of TCP-New Jersey is much better than all other TCP variants for high Random Packet Loss Rate. Results show that when we have to deal with the protocol overhead TCP-New Jersey and TCP-Vegas have lowest overhead than other TCP variants. Analyzing the impact of TCP variants on the performance of MANET routing protocols is also important to improve their performance in highly mobile networks. In this paper we have also evaluated the performance of TCP-New Jersey and TCP-Vegas with DSR, AODV and DSDV. Simulations show that MANET routing protocols perform well in combination with TCP-New Jersey than with TCP-Vegas.