Abstract:
Flash-based solid state disks (SSDs) use flash memory chips as their storage media, which have the features of non-volatility, small size, light weight, high shock resistance, high performance, and low power consumption. Single SSDs have the drawbacks of poor random write performance and limited erase endurance. Organizing multiple SSDs with the RAID technology is promising in delivering high reliability, large capacity and high performance. However, researchers have demonstrated that applying RAID algorithms to SSDs directly does not work well and have proposed some SSD-aware RAID algorithms. In this paper, we first analyze the drawbacks of flash-based solid state disks and the RAID technology, and use performance, reliability and price as the evaluation criteria of the approaches to building SSD arrays, and choose random writes, small writes, garbage collection, load balance, erase/program cycles, wear leveling and redundancy levels as the analysis metrics. Then, we analyze and compare the advantages and disadvantages of two types of array building approaches on the disk level and the flash-chip level respectively. Finally, we summarize those different approaches and point out prospective research directions in the future.