LRU-SAI The Use of LRU Algorithm with Separation of Active and Inactive Pages to Improve Solid State Storage Device Performance

In this research, we study the new cache method LRU-SAI to boost the performance of file system access. Various cache algorithms have been developed to improve the I/O performance, but most of them are focusing on minimizing the replacement cost and improving the cache hit ratio. Our new algorithm, LRU-SAI, is to make separation of active and inactive data in the cache, then write pages into the SSD block according to the activity level of these pages. Active pages will be written to active blocks and inactive one to inactive block. As a result, SSD blocks will have a bimodal distribution. This distribution can significantly reduce the cleaning overhead on SSD. We find that LRU-SAI can greatly improve the performance by approximately 20% when the cleaning efficiency in SSD is low. Even with high cleaning efficiency, LRU-SAI can work as well as CFLRU, improving the system by about 9% compared to the LRU.