Congestion avoidance using adaptive random marking

Abstract

Recent work on congestion control in TCP/IP networks combines improved end-user transmission mechanisms with active queue management schemes. In this paper we propose MARS (Marking based on Adaptive Random Scheme), an adaptive early packet marking active queue management mechanism. Unlike previous proposals, MARS adapts the marking probability to drive the average queue length towards a target size. By doing so, MARS implicitly reacts to changes in the network dynamics without requiring per-flow state or estimates thereof. Our simulations confirm MARS's flexibility and effectiveness in reducing packet loss, controlling queuing delay, and improving utilization and flow throughput.