Improving cache hit rate through cooperative caching in cluster-based web server

Cluster systems provide not only a low cost but also a flexible alternative to fault-tolerant systems used for applications that require high-throughput and high-availability. Recently, research in cluster systems is concentrated on load balancing and caching policy in contents-based cluster. Cooperative caching has also been relied upon for achieving high-performance. In this paper, we propose a method which increases cache hit rate with hint-based cooperative caching in a web server cluster that provides high-availability, high-performance, and scalability. Our proposed method combines hint-based cooperative caching and DFR mechanism. The new mechanism uses access probability of each service based on the DFR(Duplicate Copy First Replacement) and the D-Hint in order to get higher cache hit rate. We have performed a trace-driven simulation for a cluster-based web server, and compared the method that uses service access probability with DFR. The simulation results show that our mechanism has less disk access rate.