异构Web服务器集群的自适应成比例延时差异服务模型  

A Self-Adaptive Proportional Delay DifferServ Model for Heterogeneous Web Server Clusters

作　　者：杨群[1] 谭建[1] 许满武[1] 

机构地区：[1]南京大学软件新技术国家重点实验室南京大学计算机科学与技术系

出　　处：《电子学报》2007年第5期816-822,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60273035);江苏省科技攻关项目(No.BE2003064)

摘　　要：随着Internet的迅速发展,Web服务在满足数量巨大而日益增长的社会需求中所起的作用越来越大,应用对Web服务器提出了更高的QoS(Quality of Service,服务质量)要求,特别是,要求其处理数量不断变更的用户访问,它们提出大量并发请求,并具有不同的QoS需求等.如何提高Web服务器QoS性能指标已成为当今研究的一大热点问题.本文提出一种成比例延时差异服务的异构Web服务器集群模型,该模型采用[M/M/1]:[∞/∞/FCFS]队列模型刻划集群中各结点请求队列的请求到达和服务过程;在此基础上,提出两种基于反馈控制机制的请求派发自适应修正算法,每种算法均实现了相应的请求选择和派发策略.实验数据表明,本文所提出模型是适用于异构Web服务器集群的,两种请求派发自适应修正算法均能使异构Web服务器集群获得较好的服务质量(QoS)性能,具有较好的理论价值和应用前景.With the development of Internct technology, the Web Service is playing a more and more important role in satisfying the needs of large and growing community,this in tam demands that Web Servers meet the high QoS（Quality of Service） requirements of applications. In particular, Web Servers needs to deal with problems as highly concurrent requests, different clients with different QoS requirements, and so on. Therefore,how to improve the QoS of Web Servers has become a hot research topic. This paper presents a new proportional delay differserv-enabled model for heterogeneous Web Server clusters. The model describes the request arriving and serving process of cluster nodes by [ M/M/1 ] ： [∞/∞/FCFS] queuing theory. The paper also puts forward two adaptive algorithms for request selection and dispatch,both algorithms are based on feedback control and each of them has a unique request selection and dispatch strategy. The experiment result is also given. It shows that the model presented here is suitable for heterogeneous Web Server dusters, as the clusters can achieve better QoS performance when these two algorithms are applied.

关 键 词：WEB服务器集群 服务质量(QoS) 差异服务 成比例延时 请求派发 排队论 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]