An analytical performance and power model based on the transition probability for hard disks

Abstract

In order to reduce power consumption in disk-based storage systems, based on the definition of multiple power states, a lot of efforts focus on power consumption management mechanisms which can be developed to allow system software to control transitions between the power states. But, little work studies energy consumption and performance metrics of I/O systems through the transition probability among different disk modes. In this paper, we define the transition probability mathematically from active mode to idle mode, and propose a novel analytical approach to evaluate energy consumption and performance metrics, i.e., queue length, throughput, response time. Our proposed model might motivate storage researchers to exploit a quick analysis toolkit and give some insights for the design of power-aware storage systems.