Analytical models for mixed workload multimedia storage servers

Abstract

Efficient resource management is essential in providing scalability in large multimedia information systems serving a variety of applications. In order to explore design tradeoffs in such systems, what is needed are accurate analytical models of mixed workload servers with tractable solutions. Motivated by the need for mixed workload storage servers as well as the need for performance studies of such systems, we first present a family of scheduling algorithm for mixed workload storage servers. We then develop a set of corresponding analytical non-Markovian models. We show how our analytical solution methodology applies to this entire family of models and how it can be used to obtained performance measures of interest for different classes of workloads. Lastly, we illustrate through numerical examples how such models can be used to study performance tradeoffs and facilitate the making of design choices in mixed workload storage servers.