Research And Implementation Of SSD Secondary Caching Mechanism In RAID

With the blooming development of Internet, the demand of data storage for enterprise is growing exponentially. Usually the solution is RAID(Redundant Array of Independent Disks), which is increased storage functions by combining multiple disk drive components into a logical unit and reliability through redundancy. SSD(Solid-state drive) is newly developing storage devices which is quickly developing. It has the feature of high performances and low energy consumption, which make it is attractive for modern data storage. But SSD is expensive, and the price of the same capability is nearly 100 times higher than HDD. That is the reason why SSD can't completely replace the traditional hard disk. A solution of making SSD as secondary cache for the question is proposed, making the balance of proformance and price.First, in-depth study of the feature of SSD, designing caching algorithm, improving the access time and latency by using the feature of high performances and non-volatile. Use frequency/request size as a indicator to evaluate the hot blocks and find out the high-cost blocks. Meanwhile improve the LRU algorithm, using block table to record the hot blocks, and caching high-cost blocks to SSD, to improve IOPS and throughput of RAID system.Several goals from policy design and system deployment perspectives is achieved to exploit the potential of SSD. There is no intrusively change for using SSD. Our design carefully isolates complex details behind a standard block interface. Improving system performance without intrusively changing upper-level components (such as file systems). It needs only minimizes changes to existing systems and guarantees compatibility and portability.The test result of SSD as secondary cache shows that the I/O performance is improving, because the active data is caching in SSD. It is valuable in practice for many current file systems such as on-line transaction processing and office/engineering environments.