一种基于路由器队列法则的增加SRED算法  

AN ENHANCED SRED ALGORITHM BASED ON QUEUE LAW OF ROUTER

作　　者：文宏[1] 李仲宇[1] 吴海波[1] 唐玉华[2] 

机构地区：[1]湖南科技大学计算机学院,湖南湘潭411201 [2]国防科技大学计算机学院,湖南长沙410073

出　　处：《计算机应用与软件》2007年第11期23-24,46,共3页Computer Applications and Software

基　　金：国家863计划资助项目(2003AA115130);湖南省教育厅科研项目(05C184);湖南省科技大学重点教研项目(G30304)

摘　　要：主动队列管理是网络拥塞控制的主要研究方向。作为一种典型的主动队列管理算法,SRED采用一种负载相关的概率来丢弃报文,使得路由器队列长度处于一个较低的稳定值,从而获得相对稳定的网络延时。但SRED算法在设计初始丢弃概率函数时不能保证函数曲线的平稳过渡,这将导致路由器队列长度不稳定以及不必要地丢弃报文。引入路由器队列法则,对SRED算法的缺陷进行了理论分析,在此基础上设计了一种增强的SRED算法:E-SRED。仿真实验验证了E-SRED算法在保持路由器队列稳定性以及使队列快速收敛方面对SRED算法都有一定的提高。Active queue management is one of the key techniques in network congestion control. As a typical active queue management algorithm, SRED preemptively discards packets with a load - dependent dropping probability to make the router＇s buffer occupation stabilize at a low level when the buffer in router seems congested, so that comparably steady network delay can be achieved. But the probability function curve may change suddenly, which results in the instability of queue length in router and unnecessary dropping of packets. A router queue law is introduced to analyze the limitation of SRED, and an enhanced SRED algorithm E - SRED is designed. The E-SRED algorithm is implemented on NS-2.26. The experimental results validate that E-SRED algorithm can keep queue length more stable and make the queue length converge quicker than SRED.

关 键 词：主动队列管理 拥塞控制SRED 队列法则 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]