论文部分内容阅读
网络延迟问题是用户QoS的主要问题之一,它依赖诸多因素如网络带宽、传输延迟、排队延迟和客户机及服务器的处理速度。目前主要采用缓存和预取技术来减少网络延迟,但缓存技术所提高的缓存代理服务器的命中率是有限的。该文系统地阐述了目前预取算法的基本思想并把它们分成四类:基于流行度、基于交互、基于访问概率和基于数据挖掘的预取算法。在对它们进行分析比较的基础上,提出了一种智能的预取方案。该方案使用模糊匹配来计算用户对页面的访问概率,同时要控制预取的量和预取的时刻,以避免对网络的性能产生负面影响。
Network latency is one of the major issues for user QoS, which depends on factors such as network bandwidth, latency, queuing latency, and client and server processing speed. At present, the main caching and prefetching technology to reduce network latency, but caching cache technology to improve the hit rate of the proxy server is limited. This paper systematically expounds the basic concepts of current prefetching algorithms and divides them into four categories: prefetching algorithms based on popularity, based on interaction, access probability and data mining. On the basis of analyzing and comparing them, an intelligent prefetching scheme is proposed. The scheme uses fuzzy matching to calculate the probability of users accessing the page, and at the same time, the amount of prefetching and the timing of prefetching should be controlled to avoid negative impact on the performance of the network.