Using Replication for Energy Conservation in RAID Systems.
Jinoh Kim, Doron Rotem · Parallel and Distributed Processing Techniques and Applications · 2010
Energy efficiency has become a major concern in data centers as several reports predict that the anticipated energy costs over a three year period will exceed hardware acquisition. In particular, several reports indicate that storage devices (and cooling them off) may contribute over 25 percent of the total energy consumed in a data center. In this paper, we present a novel approach for energy conservation, called iRGS (inter-RAID Gear-Shift), which utilizes data replication as a tool for extending the idleness period of a large fraction of the disks. iRGS adapts to the workload observed by the system, thus allowing energy saving subject to required service level. In particular, iRGS can manage power in large data centers with multiple RAID groups unlike previous work which deals with individual disks on one RAID group. To enable this, iRGS provides (1) a new replication algorithm that allows gradual adaption to the workload by gear shifting, (2) a mapping technique to service requests under power saving modes, and (3) a write consistency mechanism for new writes and existing file updates under replicated environments. Simulation with real life trace data (Cello99) and synthetic data shows that our method saves up to 60% of energy, and outperforms existing power management algorithms, while providing better response time.