Modelling rate-based dynamic cache sharing for distributed VOD systems

Abstract

A distributed VOD system includes several VOD sub-systems, each VOD sub-system consisting of an archive server (AS), a continuous media server (CMS) and a medata DB. A VOD sub-system employs an object replacement algorithm by which a video is selected to be replaced by a new video. Upon a miss, the VOD system must decide onto which CMS to load the new video. We address this issue by modelling a dynamic approach of logically sharing the overall CMS cache space among the CMSs based on the current average arrival rate of requests at each VOD sub-system to achieve superior cache performance. Our simulation results show that dynamic cache sharing is a means by which the hit performance at the CMSs can be improved.