论文部分内容阅读
随着全球数据量的爆炸式增长,传统磁盘阵列构建的大型存储系统在性能上已经无法满足应用的需求.新兴的固态盘存储介质,具有高性能、低功耗等优良特性,将它作为二级缓存引入传统磁盘阵列可以有效改善性能瓶颈.SHCA(SSD aware Heterogeneous Cache Algorithm)算法根据磁盘阵列与固态盘的不同存储特性,将内存自适应地划分为固态盘缓冲区和阵列缓冲区.固态盘缓冲区采用一种基于缓存块淘汰代价的置换算法,根据缓存块的热度和固态盘的读写性能差异,对淘汰缓存块引起的代价进行评估,优先置换淘汰代价最低的块.阵列缓冲区采用基于条带的最近最少使用算法,利用多磁盘的并行性,提高阵列读写吞吐率;同时针对有校验信息的阵列,减少写惩罚.阵列缓冲区淘汰的数据块经过热点数据筛选,识别为热点的块缓存至固态盘,减少阵列读写次数.在详细介绍了SHCA算法设计的同时,进行了相应的试验测试和性能分析.
With the explosive growth of the global data volume, the large-scale storage system constructed by the traditional disk array has been unable to meet the application requirements in terms of performance.The emerging solid-state disk storage medium has the excellent characteristics of high performance and low power consumption, Level cache introduces the traditional disk array can effectively improve the performance bottlenecks.SHCA (SSD aware Heterogeneous Cache Algorithm) algorithm based on the different storage characteristics of the disk array and the solid state disk, the memory is adaptively divided into a solid-state disk buffer and an array buffer. The buffer adopts a permutation algorithm based on the elimination cost of the cache block, and evaluates the cost caused by the elimination of the cache block according to the difference between the popularity of the cache block and the read / write performance of the solid state disk, and preferentially replaces the block with the lowest cost. Based on the stripe least recently used algorithm, the multi-disk parallelism is used to improve the read and write throughput rate of the array, meanwhile, the write penalty is reduced for the array with check information.The data blocks eliminated by the array buffer are filtered through the hot spot data and identified as Hot block cache to SSD, reducing the number of array read and write.In a detailed description of the SHCA algorithm design When, for the corresponding experimental test and performance analysis.