A Scalable Network Area Storage with Virtualization: Modelling and Evaluation using Stochastic Reward Nets

Abstract

Modelling and analysis of storage system in data centers for availability prediction is of paramount importance. Many studies in literature proposed different architectures and techniques to enhance availability of the storage system. In this paper, we proposed to incorporate virtualization techniques on a network area storage. We used stochastic reward nets to model the system's architecture and operational scenarios. Furthermore, we investigated various measures of interests including steady state availability, downtime and downtime cost, and sensitivity of the system availability with respect to impacting parameters. The analysis results show that the proposed storage system with virtualization can obtain an acceptable level of service availability. Furthermore, the sensitivity analysis also points out complicated dependences of service availability upon system parameters. This paper presents a preliminary study to help guide the development of a scalable network area storage with virtualization in practice.