利用基于测量的WFQ实现比例区分服务模型及其性能分析  

Using Measurement-Based WFQ to Implement PDS and Its Performance Analysis

作　　者：晋晓辉[1] 李建东[1] 郭峰[1] 

机构地区：[1]西安电子科技大学ISN国家重点实验室信息科学研究所,陕西西安710071

出　　处：《电子学报》2002年第3期399-403,共5页Acta Electronica Sinica

基　　金：国家自然科学基金 (No .69872 0 2 8);教育部高校骨干教师资助计划 (No.863 31 7 0 3 0 3 99)

摘　　要：本文首先介绍了比例区分服务模型的定义以及现有的两种实现算法BPR和WTP ,然后定量地分析了WFQ算法在系统负荷和业务负载分布变化时的性能变化 ,得出了其本身具有一定的抗突发能力 .提出了一种基于测量的可变权值的WFQ调度算法 ,最后通过仿真 ,将它与BPR ,WTP和WFQ进行了比较 .仿真结果证明基于测量的WFQ在业务负载正常分布的情况下性能比WFQ更好 。In the Internet,there are two types of QoS architectures: IntServ/RSVP and DiffServ.The DiffServ is divided into absolute DiffServ and relative DiffServ.First,the model of proportional differentiated service (PDS),a typical relative DiffServ model,and its two implementation algorithms BPR (Backlog Proportional Rate) and WTP (Waiting Time Priority) are presented in this paper.Then,the performance of WFQ (Weight Fair Queuing),a typical PFQ (Packet Fair Queuing) algorithm,is analyzed and simulated.The result shows that WFQ has certain resistibility to the burst of higher priority traffic classes.WFQ has better resistibility under medium load than in heavy load,and the larger the ratio of weights of successive queues,the stronger the resistibility of WFQ.Finally,a new measurement based WFQ (M WFQ) is proposed.The performances of long term and short term of M WFQ with BPR,WTP,and WFQ are given.The results show that M WFQ has the best performance under normal traffic distribution reasonably and can be adaptive to variation of traffic load distribution.

关 键 词：比例区分服务模型 WFQ 测量 因特网 性能分析 

分 类 号：TP393.4[自动化与计算机技术—计算机应用技术]