TCP PERFORMANCE AND FAIRNESS OVER AD HOC NETWORKS

Abstract

Research on congestion control in wireless ad hoc networks has mostly been limited to enhancing TCP based mechanisms. It has been shown that in an ad hoc network with multiple TCP flows, the goodput can be significantly different among competing flows. This phenomenon is particularly evident when comparing flows of short paths to those of long paths. Most of the research in this area concentrates on improving aggregate goodput. However, fairness is a critical issue that deserves more attention. Theoretical analysis of the fairness property of a protocol is very difficult if not impossible. Thus, comprehensive and systematic simulation studies of fairness have become indispensable. Also, current implementations of TCP suffer from performance problems like flow synchronization leading to further goodput degradations of the victim. In this work, we propose a new packet scheduling approach which helps competing TCP flows to achieve consistent fairness without much aggregate goodput loss regardless of the network topology, traffic load and radio parameters. Comprehensive simulations have been carried out to compare the fairness of the proposed approach with TCP NewReno and TCP AP (TCP with Adaptive Pacing) approaches. Simulation results show that the proposed approach is superior to TCP NewReno and TCP AP approaches in consistently eliminating the extreme unfairness existing in all the simulated scenarios and congestion window size is no longer a threat to fairness. The results also reveal that the proposed approach reduces the flow synchronization prevalent with TCP NewReno and TCP AP approaches.