高效能的全球資訊網文件讀取方案

Abstract

為了提昇全球資訊網文件讀取的速度，目前廣泛採用的方式是在WWW 伺服器和用戶端之間，裝置代理快取伺服器。不過由於代理快取伺服器彼 此沒有互相結合，因此往往會發生同一份資料，被同一個區域的代理快取 伺服器重複讀取的現象。為了避免這種情形，有必要將同一區域的代理快 取伺服器加以整合，形成一個代理快取伺服器群組架構，充分運用彼此已 存有的資料。本文主要的目的，即是評估不同架構的代理快取伺服器群組 整體效能的表現。 我們共設置了三台代理快取伺服器，記錄了用戶端 使用的情形。以這些記錄來驅動我們所設計的模擬器。利用模擬的方式， 我們觀察了快取容量與擊中率的關係，限制快取中檔案的長度來分析對擊 中率的影響，以及評估六種不同的代理快取伺服器群組架構模式。最後根 據我們觀察的現象和評估的結果，建議了一種高效能的全球資訊網文件讀 取方案。 Proxy-caching is a popular way to reduce WWW document access delay and server load. Two different proxy cache servers in the same local area network may request the same document. It is beneficial to construct a scheme in which all the proxy cache servers cooperate to fetch WWW documents in a local area network so that the delay of user requests can be reduced, and the network load and server load are reduced as well. The goal of this paper is to evaluate cooperative mechanism among proxy cache servers and present an efficient WWW document retrieval scheme. Three proxy servers are set up in the local area network of our department. The client access requests on the proxy servers are all stored into trace logs. A trace-driven simulator is developed to analyze the design issues of proxy servers. We evaluate the effect of cache size on the cache hit rate, the effect of limiting the maximal file size stored in cache on the hit rate and the performance of six cooperative schemes of proxy cache servers. According to these simulation results, we propose an efficient WWW document retrieval scheme for domestic WWW users.