Data placement in intermittently available environments

Abstract

In this paper, we address the problem of data placement in a grid based multimedia environment, where the resource providers, i.e. servers, are intermittently available. The goal is to optimize the system performance by admitting maximum number of users into the system while ensuring user Quality of Service (QoS). We define and formulate various placement strategies that determine the degree of replication necessary for video objects by using a cost-based optimization procedure based on predictions of expected requests under various time-map scenarios and QoS demands. We also devise methods for dereplication of videos based on changes in popularity and server usage patterns. Our performance results indicate the benefits obtained the judicious use of dynamic placement strategies.