ANALYSIS OF CACHE REPLACEMENT POLICIES ON WORLD-WIDE-WEB

Abstract

The success of the World Wide Web has led to a steep increase in the user population as well as the amount of traffic on the internet and this tremendous growth of World Wide Web is continue to strain the internet. This growth has generated mainly three problems namely, scarce network bandwidth, high latencies and overloaded servers. Caching has been recognized as one of the most important technique to reduce high bandwidth requirement, high latencies, overloaded servers and with recent explosion in usage of World Wide Web, the problem of caching web objects has gained considerable importance. This dissertation aims at a comparative study of various caching policies along with novel caching policies designed specifically for web objects called Pyramidal and Geometrical selection schemes. Pyramidal selection scheme (PSS) combines simplicity and adaptivity. The key idea is pyramidal classification of objects depending upon their sizes and introduction of auxiliary cache, which ensures a good admission control policy. A new caching scheme based on geometrical classification of web object depending upon increasing range of access costs, expiration times and sizes of objects is proposed in the dissertation. The new scheme is known as Geometrical Selection Scheme (GSS). We use event driven simulation to compare Hit Ratio, Cost Saving and Robustness of conventional policies as well as PSS and GSS policies and observe that PSS and GSS policies have good hit ratio, cost saving and are more robust to varying workload characteristics. It has been shown that GSS is more practical and viable variant of PSS caching algorithm, it retains all advantages of PSS scheme and reduces complexity of PSS algorithm.