HALO: a Fast and Durable Disk Write Cache Using Phase Change Memory

Zhuo Liu,Bin Wang,Weikuan Yu

DOI: https://doi.org/10.1007/s10586-017-1076-0

2017-01-01

Cluster Computing

Abstract:To close the increasing performance gap between disk storage and processors, flash based solid-state devices (FSSDs) have been adopted within the memory hierarchy to improve the performance of hard disk drive (HDD) based storage system. However, FSSDs still suffer from erase-before-write restriction, coarse access granularity and limited write endurance. Recently, the cutting-edge non-volatile memory technologies are merging, e.g., phase-change memory and resistive memory, which offer us new storage alternatives with faster access, byte-addressability and better endurance. In order to address the imperative data intensive computing issues, we propose to leverage PCM as disk write cache for constructing a hybrid PCM+HDD storage architecture in this paper. First, we develop a novel hash-based write caching scheme called HALO to improve both spatial and temporal locality on hard disks, thus addressing the limitations of traditional LRU caching algorithms and rendering better I/O performance. To deal with the limited durability of PCM devices and reclaim the degraded spatial locality in previous wear-leveling techniques, we further propose novel PCM wear leveling algorithms that provide effectively uniform writes while maximizing access parallelism. We have evaluated this PCM-based hybrid storage architecture using applications with a diverse set of I/O access patterns. Our experimental results demonstrate that the HALO caching scheme leads to an average reduction of 36.8% in execution time compared to the LRU caching scheme, and that the space filling curve based wear leveling extends the lifetime of PCM by a factor of 21.6.