Hybrid Prefetching for HTTP Proxy Servers

Abstract

隨著全球資訊網上各種應用的快速成長，減少使用者等待網路內容下載時間的技術變的日益重要。在代理伺服器加上快取功能是其中最有效而且已經廣泛使用的機制之一。以代理伺服器預先抓取文件可以提高快取的成功率而讓客戶端得到更快的反應速度。本文提出一種混和式的預先抓取技術，結合不同的限制條件，可有效提高快取命中率，同時所產生的額外網路流量亦得以受到控制。我們也發現在有預先抓取機制的代理伺服器上，如果有總快取容量不大的條件下，把快取分為兩部分會有比較好的效果。As the application of World Wide Web grows rapidly, technologies to reduce web latency become important. Proxy caching is one of the most effective and widely implemented mechanisms to shorten the response time. If documents are prefetched and stored in the proxy server, higher cache hit rate and faster response time can be achieved. In this paper, we proposed a hybrid prefetching technique, which combined the content parsing and referer techniques. Cooperating with probability and popularity constraints, the hybrid prefetching increase cache hit rate significantly. In the mean time, the traffic overhead is limited. All of the mechanisms introduced in this paper are experimented by trace driven simulation.