基于参数估计的随机早期探测改进算法

流量控制是互联网有效运行的必要条件.随机早期探测RED(Random Early Detection)是一种基于整条链路的积极队列管理技术,它能显著地提高网络的利用率、稳定性和公平性,是IETF倡导使用的流量控制技术.它由Hashem首先提出雏型,Floyd对之作出了重大改进,形成了较为完善、实用的RED技术,为了在应用中增大RED的适应范围,Fang又提出基于参数自适应的RED.由于现有方法未能稳定、准确地刻画网络流量特征,在流量特性差异较大时,以上方法均难以有效控制网络流量.在本文中,作者首先给出了估计流量特性的解析方法,该方法较参数自适应具有更好的稳定性和准确性;并据此提出了一种RED改进算法,最后通过动态仿真,证明该改进算法提高了RED的适应性,降低了网络时延,减小了时延抖动.

The Improvement of RED Based on Estimating Parameter

The flow control is vital for Internet.Random Early Detection(RED) is an active queue management based on the link,which improves efficiency,stability and impartiality,and is also advocated by IETF.It is presented by Hashem,then had been improved into rather perfection and practicality by Floyd,and further had been appended with the self-configuring by Fang to be more adaptive.After simulation,we find that both RED and the self-configuring can't effectively control the network flow while confronted with large variety of flows,because they can't figure the property of network flows accurately and stably.In this paper,we give the analytic way to estimate the parameter of RED,which is more accurate and stable than the self configuring,and then present the amendatory algorithm of RED,which is demonstrated by simulation to be more adaptive,with lower transfer delay and jitter.