Dynamic data reorganization for energy savings in disk storage systems

  • Authors:

  • Ekow Otoo;Doron Rotem;Shih-Chiang Tsao

  • Affiliations:

  • Lawrence Berkeley National Laboratory, University of California, Berkeley, CA;Lawrence Berkeley National Laboratory, University of California, Berkeley, CA;Lawrence Berkeley National Laboratory, University of California, Berkeley, CA

  • Venue:
  • SSDBM'10 Proceedings of the 22nd international conference on Scientific and statistical database management
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

High performance computing (HPC) systems utilize parallel file systems that are striped over large number of disks to address the I/O performance requirement of data intensive applications. These large number of disks are typically configured into RAID groups for resiliency. Such arrangements of on-line disk storage systems constitutes one of the major consumers of power in HPC centers. Many disk power management (DPM) schemes have been suggested where by the power consumed by these disks is reduced by spinning them down after they experience long idle periods. Spinning the disks down and up results in additional energy and response time costs. For that reason, DPM schemes are effective only if the disks experience relatively long idle periods and the scheme does not introduce a severe response time penalty. In this paper we focus on RAID storage systems where by, depending on the RAID level, a group of disks are formed into RAID units. We introduce a dynamic block exchange algorithm which switches data between such units based on the observed workload such that frequently accessed blocks end up residing on a few "hot" units thus allowing the majority of RAID groups to experience longer idle periods. We validated the effectiveness of the algorithm with several real-life and synthetic traces showing power savings of up to 50% with very small response time penalties.