AIMD组播拥塞控制在卫星IP网络中的性能分析及改进

Performance Analysis and Improvement of AIMD-based Multicast Congestion Control in Satellite IP Networks

基于AIMD的组播拥塞控制算法由于采用了与TCP类似的拥塞控制策略,可以实现TCP友好性,因而在组播传输协议中得到了广泛应用。为了分析卫星网络长延时、高误码率特性对AIMD组播拥塞控制算法的影响,本文采用一种基于马尔可夫随机过程的理论模型,把拥塞发现时刻的拥塞窗口值作为马尔可夫链的状态;根据该理论模型,推导了系统吞吐量与卫星网络各种参数的关系式,从而分析了基于AIMD的组播拥塞控制算法在卫星网络中的性能。在此基础上,提出了采用接收者分组的方法来提高系统吞吐性能的改进方案,并对不同信道条件下的最优分组数量以及带来的吞吐量增益进行了研究。数学仿真结果表明,在高误码率、长延时的卫星网络中,采用最优分组可以显著提高组播系统的平均吞吐量。

Due to the TCP-like congestion control policy, AIMD-based multicast congestion control algorithms achieve TCP-friendly and are widely used in multicast transport protocols. In order to reveal the influence of long link delays and high link errors in satellite IP networks on AIMD-based multicast congestion control algorithms, a theoretical model based on Markov stochastic processes is studied, in which the congestion windows at congestion detection moment compose the states of the Markov chains. The throughput performance of AIMD-based multicast congestion control algorithms is analyzed with this theoretical model. Simulation results show that, in case that a large number of receivers are involved, the method of dividing the receivers into some groups and maintaining a different multicast session to each of the groups can achieve high throughput.