Dynamic contents clustering for disk I/O enhancement in web proxy server

Abstract

With the increasing popularity of World Wide Web, network traffic due to HTTP requests and responses has dramatically increased. World Wide Web proxies are used to reduce user-perceived latency and wide-area network traffic by caching the frequently accessed Web objects. But, It is said that the bottleneck of Web proxy cache server is disk I/O in recent researches. It grows user-perceived latency significantly. In general, proxy is running on a UNIX file system that has different disk I/O workload characteristics than those of proxies. It does slow down the performance of proxies. In this paper we study the problems of caching web objects on disk, and suggest a Dynamic Contents Clustering for disk I/O enhancement, which preserves locality of the HTTP requests by translating it into spatial locality in disk space. It reduces seek and rotation delay of disk I/O by increasing the unit of disk I/O. Additionally we use BUDDY disk layout system in a file for reduction in file meta-data management overhead and an easy disk space allocation for the cluster of Web contents. We evaluate the performance of our approach by the mix of trace-driven simulation and experimental evaluation. And, we compare our approach with SQUID proxy server, the most popular caching proxy within the public domain, and a previous work that use BUDDY file management scheme for disk I/O enhancement because our approach is an enhanced version of this previous work.