Measurement Based Intelligent Prefetch and Cache Technique in Web

In Web, users in the same workgroup might have similar interests and habits. In this paper we study intelligent proxy techniques for people who use a proxy to access Web. Our intelligent proxy has two parts: cache and prefetch. We researched three replacement policies for proxy cache: LRU and two variations of LRU. Our simulation showed that basic LRU produces worst hit rate, while the other two policies achieve roughly the same results. Therefore, we choose a mixture of them as our proposed scheme. By introducing the prediction algorithm and threshold algorithm, The proxy can predict which Web files will be needed in the near future and download some of them before they are really requested by the user group. We can get a more accurate probability by running it on the proxy server than on the client site, when client user has not visited a file in a Web server often enough. By implementing the techniques of prefetch and cache on the proxy server, we can reduce the Web latency perceived by users and also the total number of access requests to Web server, thus propose a reasonable way of decreasing their cost to access the Web for developing countries. 1