路由器算法Droptail和RED的分析与优化  

Analysing and Optimizing of the Droptail and RED Router Algorithm

作　　者：刘佳琦[1] 陈志刚[1] 李登[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机工程与应用》2006年第19期90-92,173,共4页Computer Engineering and Applications

基　　金：国家自然科学基金资助项目(编号:60573127);湖南省自然科学基金资助项目(编号:03JJY4054)

摘　　要：因特网的迅速发展使网络拥塞问题日益得到人们的重视,近年来提出了许多基于路由器的拥塞控制机制。文章详细分析了基于路由器的弃尾(Droptail)和随机早期检测RED(RandomEarlyDetection)拥塞控制机制的设计思想和算法实现。针对RED算法存在的不足对其进行改进,提出CRED(ConicRandomEarlyDetection)算法,该算法的基本思想是,当平均队列长度在最小和最大阈值之间时,使丢包概率采用一种平滑的增长机制。通过大量仿真实验对三种算法的性能进行对比研究,表明CRED算法在网络延迟、振荡等网络性能上比RED算法更有效。With the rapid growth of the Internet,mueh attention has been drawn to congestion control problems.Many network congestion control mechanisms which are used in routers have been proposed in recent years.The paper first analyses design principle and algorithm implementation of the Droptail and RED congestion control mechanisms which are used in routers.Then the paper analyses the weakness of RED and improves on the algorithm,proposing the CREI） algorithm.When the average queue length is between the min-threshold and max-threshold,the drop probalbility rises with slow speed.Then the performances of three algorithms are compared by many simulation experiments.The results indicate that the CRED algorithm can achieve better network performance than RED algorithm considering such parameters as delay,delay variation and so on.

关 键 词：网络拥塞 队列管理算法 队尾丢弃 随机早期检测 NS2 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]