自相似流量随机早期检测方法

Random Early Detection Algorithm of Self-similar Traffic

自相似流量环境下的随机早期检测(RED)算法可以充分考虑网络流量特性,提高网络拥塞控制的效率.研究了自相似流量环境下RED算法的参数设置问题,根据网络流量自相似性的特点,探讨了自相似流量环境下RED算法的参数设置方法,给出了算法的基本构架及其实现步骤.基于分形布朗运动及其包络过程,导出了自相似流量环境下RED算法最大队列长度阈值和丢包概率的计算公式.仿真结果表明,所提出的自相似流量RED算法能明显减少队列长度波动,提高链路利用率,在相同包丢失概率条件下可接纳更多的连接.

A random early detection （RED） algorithm of self-similar traffic with self-similar traffic input can improve the effectiveness of network congestion control by taking the characteristic of traffic into account. Setting of parameters for the RED algorithm was investigated. Based on the characteristic of self-similar traffic, a parameter setting scheme for the RED algorithm of self-similar traffic was proposed, and the structure and implementation process of this algorithm were given. Based on fractional Brownian motion （FBM） and its envelope process, formulas of calculating the upper queue threshold and the packet drop rate in the RED algorithm were derived. The simulation results indicate that the proposed RED algorithm can well control the queue size and improve the performance of active queue management. Compared with the standard RED algorithm, the proposed RED algorithm can admit more connections under the condition of the same packet loss probability.