关键词: 集群服务器 用户访问模式 预取 相似度 聚类分析

Abstract:

Using the user access log taken in the environment of the cluster server, we propose an implementanon ot user sessions clustering which is indispensab   le to the pre-fetching component. We develop a Perl script to generate the user session set from user access log files, and then enhance the Leader clus  tering algorithm using a new similarity metric instead of the Euclidean distance. Also, we calculate the inlernal distances and interval distances of cl  usters to evaluate the algorithm. Experimental results show that the implementation can cluster the user access logs effectively, and the temporal and s  pecial efficiency can meet the on-the-fly need of Web server pre-fetching.

Key words: （cluster server, user access pattern, pre-fetching, similarity, clustering analysis）