文摘
Big Data Analytics is a big challenge for the performance of the computing and storage systems. With the rapid development of multi-core and GPU processors, the performance of HDD-based storage system becomes much more serious. The flash-based Solid State Disks (SSDs) have become an emerging alternative to HDDs and received great attentions from both academia and industry. However, a single SSD cannot satisfy the capacity, performance and reliability requirements of a modern storage system supporting increasingly demanding data-intensive computing and applications. Redundant Array of Independent SSDs (RAIS) is an effective way to build high-performance, high-reliability, and high-capacity SSD-based storage systems. In RAIS, the chunk size is an important parameter that affects the system performance. However, the existing studies are mainly focused on the efficiency of chunk size of RAID. Because of the different performance characteristics between HDDs and SSDs, the results of these studies could not be applied to the RAIS. In this paper, we first conducted extensive experiments on the efficiency of chunk size on the RAIS performance. Based on the experimental results, we proposed a Multi-Chunk RAIS (short for MC-RAIS) to improve the performance of the SSD-based storage systems. Evaluation results show that MC-RAIS outperforms the existing fix-chunk-size SSD-based disk arrays in the I/O performance measure by more than 50 %.