Impact of frequency-based cache management policies on the performance of segment based video caching proxies

Abstract

This paper examines a novel cache management policy, which we call LRLFU and is applied to groups of blocks of the media files. This cache management policy is capable of capturing the changing popularities of the various videos by attaching a caching value to every video according to how recently and how frequently the video was requested, and decides to cache the most 'valuable' videos. Our event-driven simulations have shown that the frequency considerations we have introduced in the caching values attached to the videos (i) improve the byte-hit ratio and (ii) significantly reduce the fraction of user requests with delayed starts. © IFIP International Federation for Information Processing 2004.