Research On RAID Mechanism In Storage System Based On Flash Memory

With the rapid growth of data-intensive applications, more high-performance and reliable storage systems are required. At the same time, storage system mainly comprised by disks cannot meet the requirement anymore. Flash Memory has the advantages of small volume, low power consumption, excellent shock resistance and good performance. With Flash Memory's price decreasing and its capacity increasing, it must be more widely used in the future. RAID mechanism can be used to construct RAID storage system based on Flash Memory to increase the capacity, elevate the performance and improve the reliability, which provides a feasible way to achieve the goal of constructing storage system with high performance and good reliability.To pursue the goal of constructing high-performance and reliable storage system, we make research on the application of RAID mechanism in storage system based on Flash Memory. Firstly, we analyze and compare the two different schemes of construcing RAID system which are on the SSD level and flash chip level respectively in flash storage system and decide to use RAID mechanism based on the level of flash chips. Secondly, we propose a new RAID strategy called YA-RAID. It adopts architecture of multiple levels RAID, configuration of distributed controllers, mechanism of coarse-grained and fine-grained synthetically wear leveling. We depict the design and prototype implementation strategy in detail. Thirdly, we accomplish the design and implementation of a customizable and modular flash simulator which comprised by the FTL part and Flash Memory chips simulated by file system. Based on this simulator, some application cases are presented after the validation of the simulator. Lastly, we customize the specialized simulator with several different kinds of RAID strategies including YA-RAID. We run the traces on them and the result shows that, compared to other RAID strategies, YA-RAID strategy can guarantee the good wear leveling and boost up the performance by 30% approximately.