论文部分内容阅读
本文针对固态盘的写代价高于读代价的特性,提出了一种面向闪存固态盘的缓存管理算法FLRU.其核心思想是为每一类数据页维护一个双缓冲区LRU队列:冷数据LRU队列,热数据LRU队列.提出的算法优先置换冷的干净数据页和冷的脏页(从冷数据LRU队列中置换),其次从热数据LRU队列中置换干净数据页,最后置换热的脏数据页LRU中的数据页.实验表明,与其它算法相比,FLRU算法能够适应不同类型的读写负载,同时明显降低了读写代价.
Aiming at the fact that the write cost of SSD is higher than the cost of read, this paper proposes a cache management algorithm FLRU for flash SSDs, whose core idea is to maintain a double-buffer LRU queue for each type of data page: cold data LRU queue , Hot data LRU queue.The proposed algorithm preferentially replaces cold clean data pages and cold dirty pages (replaced from cold data LRU queues), secondarily replaces clean data pages from hot data LRU queues, and finally replaces hot dirty data pages LRU data page.Experiments show that, compared with other algorithms, FLRU algorithm can adapt to different types of read and write load, while significantly reducing the cost of reading and writing.