一种优化的闪存地址映射方法

发布时间：2018-09-14 18:36

【摘要】：近年来,NAND闪存广泛应用于各类嵌入式系统.由于"异地更新"的限制,闪存中需要地址映射方法将来自文件系统的逻辑地址转换为闪存中的物理地址.随着闪存存储空间的日益增长,如何使地址映射表占用较小的内存而又不损失较多性能,成为一个重要的问题.基于需求的页级地址映射方法能够有效地解决这个问题,然而该方法会产生地址转换页操作的额外开销,影响系统性能.从基于需求的地址映射方法出发,从两方面进行优化:首先,为了减少转换页的频繁更新,提出了页级地址映射缓存技术以统一在闪存和内存中的地址映射信息的粒度;其次,设计了基于地址转换页的数据聚集技术.通过该技术,每个数据块在垃圾回收时产生的地址转换页的更新开销被降至最低.实验用一系列基准数据集并与之前代表性的工作进行比较,结果表明,优化的地址映射方法能够大量减少额外地址转换页的开销,并提高闪存存储系统的性能.
[Abstract]:In recent years, NAND flash memory is widely used in various embedded systems. Due to the limitation of "remote update", address mapping method is required in flash memory to translate logical address from file system to physical address in flash memory. With the increasing of flash storage space, how to make address mapping table occupy less memory without losing more performance has become an important problem. The page-level address mapping method based on requirement can solve this problem effectively, however, this method will generate the extra cost of the address translation page operation and affect the system performance. Based on the requirement based address mapping method, two aspects are optimized. Firstly, in order to reduce the frequent updating of translation pages, a page-level address mapping cache technique is proposed to unify the granularity of address mapping information in flash memory and memory. The data aggregation technology based on address translation page is designed. With this technique, the update cost of address translation pages generated by each data block during garbage collection is minimized. The experimental results show that the optimized address mapping method can greatly reduce the overhead of additional address translation pages and improve the performance of flash memory system.
【作者单位】：计算机软件新技术国家重点实验室(南京大学);香港理工大学计算系;
【基金】：国家自然科学基金(61170066,61321491,61003025) 国家高技术研究发展计划(863)(2011AA010103)
【分类号】：TP333

【参考文献】