A strip level data layout strategy for heterogeneous parallel storage systems

Abstract

Parallel file systems commonly distribute a file across multiple file servers with a fixed-size strip, so that the data access can be served through multiple devices simultaneously, which is similar to RAID 0. The default data layout works well in traditional homogeneous storage systems, but as the Solid state disks (SSD) are conduct into storage system, the data layout of heterogeneous parallel storage systems need to reconsider in order to get a better performance. In this paper, we propose a strip level data layout strategy for heterogeneous parallel storage systems (SLDP), which find the optimal strip configurations for each solid state disk file server node (SNode)and mechanical hard disk drive file server node (HNode), according to data access pattern, along with the consideration of the limitation of SSD space, use variable strip size to reorganize the data layout of storage systems. the main idea is let the key regions of file get the optimal strip configuration, which can obviously improve the system performance, and rest part of the file are distributed according to the SSD free space threshold, which may reduce the system performance, SLDP can find a data layout that makes the improvement is much bigger than the reduction, so that to get totally better performance of the heterogeneous storage system. Experimental results show that the SLDP is feasible and can improve system performance.