Web集群中文档组织分布优化策略.pdfVIP

2006Vol33 11 Web * ) ( 武汉大学电子信息学院 武汉430079) Web 集群服务器已被广泛用来提高Web 服务器的性能对于如今内容海量级的大型网站来说, 如何 Web 集群服务器上组织和分布Web 文档是一个急需解决的问题本文提出了一种Web集群服务器中文档组织和分 布的优化策略,其目是减少集群系统的平均响应时间通过对Web 服务器日志的分析, 挖掘客户的访问模式,将关联 度高的网页聚类成网页簇,然后根据网页簇的负载计算其拷贝份数, 最后 集群中优化分布各网页簇的拷贝以这种 方法组织和分布文档, 可以减少服务器端的TCP 连接迁移开销, 可以实现集群内的负载均衡, 从而减小集群系统的平 均响应时间; 且相对于内容全镜像的文档分布方案, 节约了存储空间, 减少了维护各服务器文档 一致性的开销 Web 集群服务器, 网页簇,文档组织, 文档分布 Optimization Strategy for Organization and Distribution of Documents in Web Server Cluster XIONG Zhi AN PuLiu GUO ChengCheng ( Department of Electronics Information, Wuhan University, Wuhan 430079) Abstract Web server cluster has been widely used to increase the performance of Web servers. For a large website, how to organize and distribute the Web documents is imperative. This paper proposes a strategy to organize and distrib ute Web documents in Web server cluster, whose aim is to reduce the average response time of cluster system. Through the analysis of Web servers logs, discover the client s access patterns, then group the webpages closely re lated into webpage clusters, then calculate every webpage cluster s copy number according to their load, finally distrib ute the copies of webpage clusters among servers. Such strategy can reduce the overhead of TCP connection migration, and achieve load balancing among servers, thereby reducing the average response time of cluster system. In addition, compared with the method of mirror image, our strategy saves storage space, and reduces the overhead of maintaining the coherence of the documents. Keywords Web server cluster, Webpage cluster,Document organization, Document distribution , 1