Design and evaluation of an adaptive write buffer cache for solid state drives

摘要

Solid State Drives (SSD's) have shown promise to be a candidate to replace traditional hard disk drives. The benefits of SSD's over HDD's include better durability, higher performance, and lower power consumption, but due to certain physical characteristics of NAND flash, which comprise SSD's, there are some challenging areas of improvement and further research. We focus on the layout and management of the small amount of RAM that serves as a cache between the SSD and the system that uses it. Of the techniques that have been previously proposed to manage this cache, we identified several sources of inefficient cache space management due to the way pages are clustered in blocks and the limited replacement policy. We found that in many traces hot pages reside in otherwise cold blocks, and that the spatial locality of most clusters can be fully exploited in a limited time period, so we developed a hybrid page/block architecture along with an advanced replacement policy, called BPAC, or Block-Page Adaptive Cache, to exploit both temporal and spatial locality.;The idea was to adaptively partition SSD's on-disk cache to separately hold pages with high temporal locality in a page list and clusters of pages with low temporal but high spatial locality in a block list. We have developed a novel mechanism for flash-based SSD to characterize the spatial locality of the disk I/O workload and an approach to dynamically identify the set of low spatial locality clusters.;In order to evaluate the design of BPAC, we adopted the method of trace-driven simulation: we have developed a cache simulator as well as a BAST FTL simulator, in C++ language; for the purpose of comparison, the most popular flash-aware cache schemes, BPLRU, FAB, and CLC, are simulated along with BPAC; the traces we choose are collected by other researchers on OLTP servers. We ran simulation experiments and found that BPAC can effectively reduce the number of buffer replacements and so as to lower the overhead of FTL's garbage collection processes, which is a major cause of SSD's write latency; we also found that BPAC outperforms the other schemes under most cases.