Study On Server-Side Web Pre-fetching Based On Data Mining

WWW's exponential growth in size results in network congestion and server overloading. Web caching has been recognized as one of the effective schemes to alleviate the service bottleneck and reduce the network traffic, thereby minimize the user's access latency. But studies show the performance of web caching is very limited. In order to improve the performance of web caching system, web pre-fetching was put forward. Web pre-fetching mainly capitalizes on temporal and spatial locality of users' requests. When the user's current document request is loaded, his or her future document requests are anticipated by using his or her access patterns and these documents are pre-loaded or pre-fetched into a local cache. If the following document requsted by the user has been pre-loaded or pre-fetched in the cache, it can be retrieved quickly, thereby the use's access latency is reduced and the quality of web service is improved.According to the position where the pre-fetching engine resides in the WWW, web pre-fetching can be classified into client-side pre-fetching mechanism, proxy-side pre-fetching mechanism and server-side mechanism. This paper mainly focuses on the research on the server-side pre-fetching mechanism.Firstly, the idea of pre-fetching and its feasibility are introduced. The existing pre-fetching models such as pages-correlated pre-fetching, documents-clustered pre-fetching, text-matched pre-fetching and interactive pre-fetching are analyzed, and their advantages and disadvantages are listed. Three important factors must be taken into account when a pre-fetching scheme is designed. The three factors are as follows:(1) The information used to make prediction must be fresh.(2) The predictive algorithm must be efficient.(3) The ratio of performance to cost must be reasonable.Secondly, on the basis of above three factors, aiming at the peculiarities and difficulties of server-side pre-fetching mechanism, a new pre-fetching scheme, pre-fetching for all the users based on the current request, is put forward. The scheme capitalizes on the temporal and spatial locality of users' requests. With the means of association rules mining, web pages' correlation in web users' requests is discovered and used to anticipate users' future documents requests. The scheme belongs to pages-correlated pre-fetching model. By comparison with the general pages-correlated pre-fetching scheme, the log file explored is latest, the spatial complexity is less,...