大时滞网络中的拥塞控制算法

Congestion Control Algorithm in Large-Delay Networks

主动队列管理(AQM)通过网络中间节点有目的的分组丢弃实现了较低的排队延时和较高的有效吞吐量,是近年来TCP端到端拥塞控制的一个研究热点.已有的大多数AQM算法在设计过程中都没有充分考虑到大时滞对算法性能的影响.首先通过仿真试验证实了已有的几种典型算法控制的队列在大时滞网络中无一例外地出现了剧烈的振荡,导致瓶颈链路利用率下降和延时抖动加剧.为此,在进行了适当模型拟合处理的基础上,应用控制理论中的内模补偿原理设计了鲁棒的延时补偿主动队列管理(delay compensation-active queue management,简称DC-AQM)算法,克服了大时滞给队列稳定性造成的不利影响.仿真实验结果表明,新算法在大时滞小期望队列长度的网络配置中表现出的综合性能明显优于已有的算法,链路利用率是其他算法的3～4倍.

AQM (active queue management) can maintain the smaller queueing 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 networks are appeared, which decreases the utilization of the bottleneck link and introduces in large 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.