Dynamic Semantic LFU Policy with Victim tracer (DSLV): A Customizing Technique for Client Cache

Abstract

This work proposes an augmented facility to the existing LFU-based web cache replacement policy. The proposed policy takes into account three additional factors such as dynamicity of the cached pages, semantic relation between the cached and the incoming pages, and least frequently used web pages and the impact of the victimized pages to decide the pages to be purged. Hence, it is named as DSLV to reveal this strategy. This policy is intended to improve the performance of only the client cache. By introducing dynamicity levels, obsolete pages are filtered from the caching process. The semantic module identifies the semantic relationship between the cached pages. Therefore, the new incoming page and the cached pages can be customized as per the interest of the users and their access pattern. The best variation in LFU policy that avoids cache pollution is adopted. Victims are the pages that are removed from the cache for want of space. These victimized pages are monitored so that the victim module identifies the reason for getting victimized and tries to tune the policy for better performance measures. Simulation results ranked by the performance metrics namely file hit ratio, speedup, delay saving ratio, and number of evictions disclose strong induction for this kind of customized technique for client cache as an efficient enhancement to the already established LFU replacement policy.