A Novel ECN-Based Congestion Control and Avoidance Algorithm with Forecasting and Verifying

Hui-shan Liu, Ke Xu, Mingwei Xu

Abstract: The congestion control and avoidance algorithms can be divided into source and link algorithms. The source algorithms are based on ’pipe’ model; which includes TCP Tahoe, TCP Reno, TCP NewReno, SACK and TCP Vegas. In this paper, we present a novel congestion control source algorithm-FAV(Forecast And Verify), which measures packets delay to forecast whether there will be congestion in network, and uses ECN to verify whether the forecast is precise or not. Experiments are performed by ns2 to compare FAV with TCP Reno and TCP Vegas, and the results imply that FAV algorithm can decrease loss rate and link delay efficiently, and keep link utilization high. We think FAV can control the congestion better than other source algorithms.