Rate Smoothness and TCP-Friendliness through Two-step AIMD (TAIMD)

Abstract

With the proliferation of Internet and the increasing amount of multimedia traffic delivered over it, there has been a pressing need for new transmission controls that facilitate high smoothness and fast convergence to fairness. Achieving these objectives in end-to-end scenario is promising but challenging. In this paper, we present a novel end-to-end TCP-Friendly algorithm that uses two-step additive increase and multiplicative decrease (TAIMD) to control the congestion window using a history-based fair-share estimation. TAIMD smoothes the rate variations by using a two-step increase function and adopting a timer-based packet scheduling mechanism. Convergence performance is improved by using a second fair-share estimation based on packet loss, yet keeping the control less sensitive to the inherent short term fluctuations in the loss-based fair-share estimation. Analytically, we derive parametric relationships for TCP-Friendliness and for fairness convergence of TAIMD. Through simulations conducted using the NS-2 network simulator, performance of TAIMD is observed and compared with the performance of TFRC