The energy consumed by the high-performance storage subsystems increases the cost of operating computing systems for I/O intensive applications, both directly and indirectly. The development of energy-efficient storage systems can greatly reduce the cost of acquiring and operating large computing platforms, thereby making them cost-effective for a broad range of computational problems. Solutions need to be built as extensions to existing storage-system technologies such as RAID. This research addresses the above-mentioned problems by designing and implementing redundancy-aware, dynamic I/O scheduling algorithms, cache management software and static block distribution algorithms that serve as a foundation for building energy-efficient high-performance storage arrays for demanding I/O-intensive applications. The inherent redundancy of storage array is used to change the disk access pattern toward energy-efficiency with high-performance guarantee by power-aware request dispatch/transformation schemes. Two commonly used storage array system architectures, mirrored disk arrays and parity disk arrays are thoroughly studied. An extended variant of RAID-hybrid storage array is investigated as well. The significance of the proposed research is to solve the energy conservation problem for ever-important high-performance storage subsystems. It will have broad impact in the low-power computing and I/O systems community by providing an energy-efficient high-performance storage solution, which modern storage servers, storage clusters and data centers can benefit. During the course of the project, prototype versions of the energy efficient RAID and its extended variant-hybrid storage array will be delivered to the computer systems community.
