TLA: A Traffic Load Adaptive Congestion Control Algorithm for TCP/AQM Networks

Abstract

TCP combined with active queue management algorithm is the primary solution of the congestion control problem of IP network. Red and blue are the famous AQM algorithms but neither of them performs well when traffic load is heavy and when the traffic load changes. This paper proposes a new traffic load adaptive AQM algorithm named TLA. The objective of TLA is to stabilize the queue size m a wide variety of traffic scenarios. TLA considers not only the queue size but also the traffic load as the congestion indicator. So its drop probability can adapt to the changes of the traffic load and the congestion notification can be sent to sufficient TCP sources to mitigate the congestion level. Simulation results indicate that TLA can effectively stabilize the queue occupation independent of the number of active TCP connections thus resulting in a more predictable packet delay in the network