Employing dual-block correlations to reduce the energy consumption of disk drives

Abstract

Achieving high energy efficiency is becoming a very important challenge in designing future storage systems. This paper proposes to incorporate block correlations to perform energy-aware scheduling, thereby significantly improving energy efficiency of disk systems. Since the probability that an access to a file A will be followed by the same file (e.g. file B) that followed the last access to A is very high, we can infer the block correlations between files A and B. For example, block X of file A and block Y of file B may correlate with each other. We discover that energy-saving opportunities occur when the time interval between two accesses of block X and block Y is sufficiently large, and there are only a few sporadic requests distributed between these two accesses. Thus, when there is one access to block X, we can switch the disk drive to the low power state, properly arrange the sporadic requests, and transfer the disk back to the active state to serve the block access Y when saved energy is larger than energy penalty. Our experimental results demonstrate that the proposed energy-aware disk scheduler can strike a good balance between the energy saving, system performance, and reliability.