摘要
针对同一个工作组中成员可能对 Web有相似的兴趣点和访问习惯问题 ,研究了主要包括缓存和预取两个部分的智能代理技术 ;研究了代理缓存三种替换算法 :L RU和 L RU的两种变种 .仿真试验表明 ,L RU算法的命中率极差 ,将 L RU的两种变种相结合则是一种较好的方案 .通过引入预测算法和门限算法 ,代理服务器可以预测最近将要访问到的页面 ,并在客户实际请求提出之前有选择地下载 .因为单个客户访问某个页面的历史次数往往不够多 ,将预测算法放在代理服务器上比放在客户端的预测概率更准确 .通过在代理服务器上实现缓存和预取技术 ,可减少用户访问 Web的响应时间 ,还可减少实际访问 Web服务器的总次数 .并提出了一个为发展中国家用户缩减
In a Web the users in the same workgroup may have similar interests and habits.This paper studies the intelligent proxy techniques for people who have access to the Web.The intelligent proxy has two parts:cache and prefetch.Three replacement policies for proxy cache have been studied:LRU and two variations of LRU.The simulation shows that basic LRU gives the poorest hit rate,and therefore,a mixture of the other two policies is closen as the proposed scheme.The introduction of the prediction algorithm and threshold algorithm shows that the proxy can predict which Web files will be needed in the near future,and some of them should be downloaded before they are really requested by the user group.A more accurate probability can be obtained by running it on the proxy server rather than on the client site,when the number of the visit of a file by single clients is not numerous enough.Performing the techniques of prefetch and cache on the proxy server can reduce not only the latency of users but also the total number of access requests to the Web server.Thus,this paper provides a way to reduce the cost of access to the Web.
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2001年第5期563-567,共5页
Journal of Tianjin University：Science and Technology
基金
国家自然科学基金资助项目 (69672 0 31 )
