Write optimization strategy in S-RAID based on sequential data characteristics

S-RAID, a power-aware RAID, saves energy by grouping disks in the array and shutting down most of them. In order to avoid spinning up disks that are put into standby mode, all write operations are read-modify-write in S-RAID, and that degrades its write performance. A new data layout based on S-RAID: LS-RAID, is proposed in this paper in order to optimize the write performance of the storage system with the same power consumption. LS-RAID is suitable for sequential data storage. Each disk in LS-RAID is divided into two parts. Random I/O is kept separate from sequential I/O. That can reduce the negative impact on the storage system caused by sequential I/O. The parity calculates with a new algorithm: Data Incremental Parity algorithm (DIP). DIP calculates the parity only with the blocks that be written instead of all of them. The old data need not be read while the new data is written. That degraded write penalty caused by read-modify-write. Experiment shows that write performance of LS-RAID increased at least 56% than S-RAID, and that makes it possible for LS-RAID to save more energy with the fixed data transfer rate.