基于网络性能的智能Web加速技术——缓存与预取

Web业务在网络业务中占有很大比重，在无法扩大网络带宽时，需要采取一定技术合理利用带宽，改善网络性能。研究了基于RTT（round trip time)等网络性能指标的Web智能加速技术，在对Web代理服务器上的业务进行分析和对网络RTT进行测量分析的基础上，提出了智能预取控制技术及新的缓存（cache）替换方法。对新算法的仿真研究表明，该方法提高了缓存的命中率。研究表明预取技术在不明显增加网络负荷的前提下，提高了业务的响应速度，有效地改进了Web访问性能。

INTELLIGENT PREFETCH AND CACHE TECHNIQUES BASED ON NETWORK PERFORMANCE

Because the bandwidth of WAN is always the bottleneck of networks, in order to reduce the user perceived latency, caching and prefetching should be introduced. Studied in this paper are intelligent prefetch techniques suitable for people who have similar interests and use a common division proxy to access the Web. Based on analyzing the proxy traffic and measurement of network RTT , a new method to control the prefetching is proposed. The prefetch algorithm can predict which Web files will be needed in the near future and download some of them before they are requested by users. The prediction algorithm estimates the probability that each Web file will be requested in the near future according to the statistics characteristic of group users' visited Web history. On the other hand, a new algorithm for cache replacement is also proposed. This new method not only uses the traditional method on the cache replacement algorithm of the general proxies, which usually use simple FIFO or pure LRU (least recently used) algorithm, but also considers the cost of transfer of files. The simulation results show that this cache replacement algorithm has a higher hit rate and the new control method decreases the wait time of web access requests. These intelligent speed up techniques not only reduce the users access latency, but also reduce the total number of access requests to Web servers and network communication cost.