OAFTL:一种面向企业级应用的高效闪存转换层处理策略

OAFTL:An Efficient Flash Translation Layer for Enterprise Application

基于NAND闪存的存储设备通过引入闪存转换层来对闪存芯片进行封装,使得闪存存储设备像普通块设备一样使用.闪存转换层算法的性能很大程度上决定了闪存设备的存储性能,已有方法尽管可以在嵌入式环境下正常工作,但当应用到随机访问频繁的企业级应用环境中时存在访问性能低的问题.提出了一种面向企业级应用的闪存转换层算法OAFTL,该算法基于页级地址映射,根据访问操作的类型来组织映射项信息,通过为映射页保留日志信息来缓冲频繁修改的映射信息,以提高闪存读、写性能.实验结果表明,提出的OAFTL算法能够有效地适应企业级工作负载,同已有方法相比,综合读写性能提升了20%以上.

NAND flash based devices usually introduce a software firmware called flash translation layer （FTL） to simulate the flash memory like a block device. FTL is critical to the performance of flash-based devices. Most existing FTL algorithms work normally in embedded systems. However, they behave poorly when there are frequent random accesses in enterprise applications. In this paper, we propose an operation aware flash translation layer （OAFTL） for enterprise-scale storage devices based on page-level mapping scheme. OAFTL manages the entries cached in RAM according to read/ write operations separately. Besides, OAFTL supplies a log page for translation page to relieve frequent updates of translation information to improve performance. The experiment result shows that our OAFTL algorithm works effectively for enterprise workload. In our experiments, OAFTL improves the total performance by more than 20 percent compared with the existing methods.