A Study Of Web Cache Replacement Strategies And Prefecthing

With the rapid development of network technology and the exponential growth rate of web information and users, internet has become part of our everyday life. However, because the network bandwidth is limited, the user perceived access latency and the quality of service of the network have become an urgent problem. Thus, Web caching and prefetching have been recognized as the effective schemes to alleviate the server bottleneck , reduce the network traffic, and minimize customers'access latencies.In this thesis, the basic concepts, key technologies, classification , structure and mechanism of web cache and web prefetching systems are described. Then, a new web cache replacement algorithm, based on users'interest and access characteristics, is proposed and a prediction model, through analyzing users'access logs, is build, which allowed us to integrate web cache and web prefitching to improve network performance.The main research and contributions can be described as follows:(1) Most of web cache replacement policies do not take full advantage of users'interest and the access characteristics, and the hit rate is less than 40%. Then based on the GDSF algorithm and joined users'interest, content types and the access characteristics of the target of web, a new replacement algorithm named GDSF-AI (Greedy-Dual-Size-Frequency-Access-Interest) is proposed. Through simulation experiments in a IPTV system and compared with traditional algorithms, this strategy has better network performance, where hit rate can reach 52% and byte hit rate can reach 41.5%.(2) At present, web caching and prefetching technology research only emphasized on the improvements of replacement policy and prefetching algorithm, lack of the research is the combination of replacement policy and prefetching algorithm. In the thesis, according to the user's interest and the access characteristics, we propose the Interest-Top prediction algorithm. Then, combined with the Web cache replacement algorithm GDSF-AI and web prefetching, a new strategy named P-GDSF-AI is proposed. Through simulation experiments in a IPTV system and compared with traditional algorithms, this strategy has better network performance, where hit rate can reach 56.5% and byte hit rate can reach 45.6%. We should pay attention to that our simulation experiments is based on the"Download-Play"Set Top Box system in SiTV. Using the cache hit rate and byte hit rate as the evaluation of standards, the algorithm GDSF-AI and the strategy P-GDSF-AI presented above are compared with other traditional algorithms LRU, SIZE and GDSF. Then, the results show that joined users'interest, content types and the access characteristics, the algorithm GDSF-AI and the strategy P-GDSF-AI can improve the IPTV cache system performance effectively.