一种基于分类策略的聚簇页级闪存转换层算法

发布时间：2018-11-15 15:10

【摘要】：提出一种基于分类策略的聚簇页级闪存转换层算法——CPFTL.1)CPFTL将地址映射缓存分为热映射表缓存、冷映射表缓存和连续映射表缓存,分别用来缓存访问频繁的请求的映射项、访问不频繁的请求的映射项和高空间本地性的连续请求的映射项,有效提升各类请求的处理能力;2)为利用连续请求的空间本地性,CPFTL的连续映射表缓存预取多个连续的映射项,提高它对连续请求的响应性能;3)为减少页级映射算法的转换页读写开销,CPFTL的冷映射表缓存采用聚簇策略,即将属于同一转换页中的映射项进行聚簇,按簇进行LRU管理,当冷映射表缓存满时,根据簇的映射项个数和LRU选取合适的簇剔除到闪存.实验结果显示,相比经典的页级DFTL算法和最新的SDFTL算法,CPFTL的缓存命中率、平均响应时间、地址转换页操作次数和闪存块擦除次数都有显著提升.
[Abstract]:A clustering page-level flash memory translation layer algorithm, CPFTL.1) based on classification strategy, is proposed. The address mapping cache is divided into hot map cache, cold map table cache and continuous mapping table cache by CPFTL. It is used to cache the mapping items of frequent requests, the mapping items of infrequent requests and the mapping items of continuous requests with high spatial nativeness, respectively, so as to improve the processing ability of all kinds of requests. 2) in order to make use of the spatial nativeness of continuous requests, CPFTL's continuous mapping table buffers several consecutive mapping items to improve its response performance to continuous requests; 3) in order to reduce the overhead of the conversion page reading and writing of the page-level mapping algorithm, the cold map table cache of CPFTL adopts the clustering strategy, that is, the mapping items in the same conversion page are clustered, and the LRU is managed by the cluster. When the cold map table is full, the cold map table cache is full. Select suitable cluster to flash memory according to the number of mapping items and LRU. The experimental results show that compared with the classical page-level DFTL algorithm and the latest SDFTL algorithm, the cache hit ratio, average response time, address translation page operation times and flash block erasure times of CPFTL are significantly improved.
【作者单位】：杭州电子科技大学通信工程学院;浙江万里学院科研部;
【基金】：国家自然科学基金项目(61100044) 浙江省科技创新基金项目(2013TD03) 浙江省科技计划资助项目(2013C31100)~~
【分类号】：TP333

【参考文献】