Workload Based Optimization Model for Parallel Disk Systems

Researchers are struggling for building high reliable and energy efficient parallel disk systems for years. With the times of high energy cost, conserving energy in parallel disk systems has large impact on reducing cost of disk systems and other related backup systems. In addition, it will conserve a lot of energy consumed by parallel disks in high performance computing systems. Moreover, Disk reliability is also significant for data integrity. While different reliability models and energy conservation techniques have been developed for parallel disk systems, most solutions do not consider two factors together in an optimization prospective and they do not consider the impact of disk workload. In this paper, we build a high reliable and energy-efficient parallel disk system using Markov chains. Specifically, we develop a quantitative reliability model for energy-efficient parallel disk systems which considers the impact of disk workload. With the new model, we can achieve the optimization between disk reliability and energy efficiency in parallel disk systems.