Study On Performance Optimization Based On Web Caching And Prefetching Technique

With the rapid development of Internet network, Internet has been applied in many fields,such as industrial, military, medical, education and so on. And it brings great convenience topeople. At the same time, the number of network users rapidly expands. However, the upgradeof network hardware, such as network bandwidth, is very limited. When browsing the netsources, users usually perceive large access delay. Therefore, how to reduce users accesslatency and improve users access speed is a practical problem. By saving the contents usershave already visited, Web caching effectively reduces the waiting time for acquiring netsources. However it can not actively save contents would be visited. Prefetching techniquemakes up for the shortcoming of caching technique. And it has become an important methodof optimizing the performance of cache system and reducing the user access latency time.Prefetching technique can predict objects which users are going to visit. Therefore this papermakes use of the advantage of prefetching technique. Based on the prediction model, a cachereplacement algorithm is proposed. The simulation result shows the superiority of theproposed algorithm. Therefore the proposed algorithm can improve the performance of thecache system in some degree. The main research work of this paper can be described asfollows:1. This paper introduces the main part of the structure of Web caching system and thebasic theory of prefetching technique. Then the different methods of caching and prefetchingare introduced. Four main characteristics of Web access are described in detail. Research ofcache replacement algorithms and prefetching algorithms are summarized in this paper.2. In order to make sure the log data is correct and available, this paper introduces theconcept of Web log mining, and the four main steps of log data preprocessing. The definitionof Markov chain is introduced, and multi-Markov chains model is described in detail. In orderto construct multi-Markov chains model, merger of class Markov chain, cluster similarity andcriterion function are discussed.3. Based on the analysis of factors affecting the cache replacement, a target function tocalculate cache object key is constructed. It is used to measure the value of objects storing incache. Using the advantage of prefetching technique, which can predict users access objects,this paper constructs a multi-Markov chain model for users. To improve the cache hit rate, a cache replacement algorithm based on multi-Markov chains prediction model is proposed.The proposed algorithm preserves objects which have great value storing in the cache and theobjects which are predicted by the model. Using the method of trace driven, simulations arecarred out. Using different performance evaluation criterias, the proposed algorithm and otheralgorithms are compared. Experimental results show that, the proposed algorithm outperformsother algorithms, and it reduces the access delay more. And then it optimizes the cacheperformance.