Reduction of web latency: An integrated proxy prefetch-cache system framework

Abstract

Present Web infrastructure is severely suffering from heavy congestion and enormous traffic burden upon the widely distributed client network around the globe due to proliferating growth in terms of overpopulated users along with day-by-day addition of latest complex and resource hungry Web applications, tools, and services. This leads to continuous increase in Web access delay (latency) faced by the users challenging the unprecedented popularity of the World Wide Web (WWW). Proxy caching is a renowned and current strategy for refining the distributed Web-based system performance particularly applicable for remote areas which severely suffers from poor network communication and limited bandwidth availability. Web prefetching is another very useful strategy which prefetches and keeps the closely related Web objects to be accessed by the users’ in future into cache, based on their interest pattern. But still, there is a significant performance gap between the users’ requests and the result provided. The intelligent and sophisticated integration of these two strategies along with one more performance enhancement technique, i.e., Web log mining, can provide an excellent and most successful solution to minimize the problem of latency and improve the overall Web quality of service. This paper provides an Integrated Proxy Prefetch-Cache System framework with the above benefits to satisfy maximum clients’ requests with improved response time as well as with minimum latency. Also, the framework has been successfully incorporated into the highly scalable, hierarchical as well as distributed geographical region-based proxy server clusters with enhanced content co-operation, efficient metadata management, and low access delay. Using real-time Web proxy server network deployment with online trace-driven results analysis, the current framework guarantees improved object consistency between clustered proxy server caches and original Web servers.