Www Access Characteristics Of The Proxy Cache

Because WWW traffic have take up the most part in Internet access, WWW access characteristics research become an import way to improve network utilization.In this paper, firstly we study the WWW access characteristics, include exploring the essence of a Web law, Zipf's law, and analyzing the logs of a Web server and two proxy servers, expound the phsical characteristics of WWW access roundly. Based on this analysis, we modify the current self-similar model of WWW access, made it reflect the characteristic we found. According to the access characteristics we found by analyzing access logs , we proposed a cache replacement policy,called Predict Size policy, which can predict future access to the document. Finally we implement a proxy cache cluster which use the replacement policy we proposed.The contribution in this paper is as follows.More study is focused on parameter ? of Zipf-like model and a fact is found: ? is related to the boundary of Web pages access frequency increase and decrease. ? and N (Web page total number) can be used to determine a unique value n. When ? increase, the first n web pages will be accessed more frequently then ever. Among n web pages, access probability increment will decrease from the first web page to nth web page. Web pages above n will be accessed less frequently then ever.We analysed a Web server log and two proxy server logs. Analysis mainly focus on Web page request distribution, Web static document size distribution, Web static document access distance distribution, Web static doment access intertime distribution. Results are compared with literatures. The difference can be considered as the access characteristics belong to our country.According to the analysis to logs, the intertime distrubiton of traffic sources does not comply to power distribution with exponentβ∈(2,3),...