Study On Web Page Prefetch At Proxy Server

With the fast growth of network, the numbers using the network increasing rapidly too, meanwhile, the development of multi-level network's application additionally made a great deal of data transport through network. Although the speed of network has been improved considerably in recent years, but because more and more users join Internet as well as the Web service and network delay, the network become more traffic, so as to people always suffer from long time waiting when they visit the World Wide Web. It will fall short of practice if we solve the problem only depend on improving the network hardware. Web prefetching is one of the most popular and effective strategies for reducing the perceived access delay, so it has drawn extensive research attention recently.The Web caching and prefetching at proxy server can reduce the user's perceived latency, it make effective use of network resource as well. It has become one of the important task how we can get better cache utilizing rate and hit rate at proxy server.In this thesis, we study the Web prefetching technology at proxy server. It's purpose is to reduce the Web latency and provide Web prefetching service of individualization for users. Our work is mainly in the following four aspects:Firstly, we analyze all data items in the access log in order to find some rules when we preprocessing the access log by transcendental knowledge. Through this way, we can filter the log record which helpless to analyze, and the calculatingSecondly, we produce two dimensions Portal-aware hot rank list by using association rules. The Portal-aware prefetching list can mine users' access behavior in Web site, so that the hot rank list can reflect users' real access behavior more essentially.Thirdly, the page information value concept has been used in prefetching methods. We consider the factor of page size, by calculating the value of page information, the prefetching list become more representative, and the caching space can be used more reasonable.Fourthly, the hierarchical clustering method has been used to find out the time-based users' access behavior. We take Web temporality into account for constructing prefetching list, based on which we predict the users' behavior. In addition to maintain the Web page fresh, this method avoid prefetching outdated page and improve cache hit rate as well.