WAPFTL:支持预测机制的负载自适应闪存转换层算法

基于NAND Flash的固态盘凭借其低延迟、低功耗、高可靠性等优点，已经开始应用于企业级服务器和高性能计算领域。针对固态盘相对较差的写性能及使用寿命有限等不足，提出了一种闪存转换层中基于页映射机制的自适应地址映射算法WAPFTL。该算法能够在地址转换过程中预测负载读写特性并自适应地调整地址映射信息缓存的策略。实验结果表明，WAPFTL能够高效协同利用负载的时间局部性和空间局部性，提高地址映射命中率，减少因地址映射而引起的额外写操作次数;同时，有效减少了垃圾回收次数，提高了SSD整体性能。

WAPFTL:A workload adaptive page-level flash translation layer with prediction

NAND Flash based Solid State Drive (SSD) has been utilized in enterprise servers and high performance computing environments due to its various advantages, such as fast access speed, low power consumption, high reliability, etc. Since SSD suffers from poor performance of random writes and limited lifespan, a workload adaptive page level address mapping algorithm, named WAPFTL, is proposed, which can predict the characteristics of incoming requests and dynamically adapt the cache strategy of address mapping information. Experimental results show that WAPFTL can improve the cache hit ratio and reduce the number of expensive extra read/write operations by jointly exploiting the temporal and spatial localities of workloads efficiently. Besides, WAPFTL can also reduce the number of erase operations, thus improving the overall performance of SSD.