Improving TCP’s Qos using end-to-end delay based congestion control techniques

Abstract

Transmission control protocol is oriented for reliable data transfer which carry more than 90% of load in wireless network. One of the key features which make TCP, implements a window based flow control mechanism. After each successful and unsuccessful delivery, it allows window to shrink and expand. A number of research’s and investigation has been done in this area to control the expansion of TCP window for better congestion control and flow control. In early days, TCP research shows that it works unsatisfactorily in rapid network and low round trip time networks when large data is sent. TCP is not appropriate for the time of low capacity and less delay networks. In this paper, we describe a delay-based end-to-end congestion control algorithm, called TCP-EXT, which is an extension of delay and acknowledge based TCP protocols like Linux,Reno, Newreno etc. The idea behind this extension is to frequently estimate the round-trip-time, which eliminates throughput and queue oscillation when round-trip time oscillates. In this context, we introduce a simulation and an evaluation of the suggested scheme, by comparing our TCP-EXT prototype with various TCP variants in terms of throughput, fairness, stability, RTT and adaptively to changes in network.
Keywords-TCP; RTT; end to end delay; congestion control; TCP-EXT.