Congestion Control Algorithm in Large-Delay Networks

TL;DR: Simulation experimental results show that the integrated performance of the proposed AQM algorithm is obviously superior to that of the existed schemes when the network configuration parameters are large delay and small queue length, and the link utilization increases 3~4 times.

Abstract: AQM (active queue management) can maintain the smaller queuing delay and higher throughput by the purposefully dropping the packets at the intermediate nodes. It is a hotspot in the current researches about TCP end-to-end congestion control. Almost all the existed algorithms neglect the impact on performance caused by large delay. In this study, firstly a fact through simulation experiments is verified, which is the queues controlled by several typical AQM algorithms, including RED, PI controller and REM, the dramatic oscillations in large delay networks are appeared, which decreases the utilization of the bottleneck link and introduces the avoidable delay jitter. After some appropriate model approximation, a robust AQM algorithm applying the principle of internal mode compensation in control theory is designed. The new algorithm restricts the negative impact on the queue stability caused by the large delay. The simulation experimental results show that the integrated performance of the proposed algorithm is obviously superior to that of the existed schemes when the network configuration parameters are large delay and small queue length, and the link utilization increases 3~4 times.