Performance impact and interplay of SSD parallelism through advanced commands, allocation strategy and data granularity

  • Authors:

  • Yang Hu;Hong Jiang;Dan Feng;Lei Tian;Hao Luo;Shuping Zhang

  • Affiliations:

  • Huazhong University of Science and Technology, Wuhan, China;University of Nebraska-Lincoln, Lincoln, NE, USA;Huazhong University of Science and Technology, Wuhan, China;Huazhong University of Science and Technology, Wuhan, China;Huazhong University of Science and Technology, Wuhan, China;Beijing Institute of Computer Technology and Application, Beijing, China

  • Venue:
  • Proceedings of the international conference on Supercomputing
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

With the development of the NAND-Flash technology, NAND-Flash based Solid-State Disk (SSD) has been attracting a great deal of attention from both industry and academia. While a range of SSD research topics, from interface techniques to buffer management and Flash Translation Layer (FTL), from performance to endurance and energy efficiency, have been extensively studied in the literature, the SSD being studied was by and large treated as a grey or black box in that many of the internal features such as advanced commands, physical-page allocation schemes and data granularity are hidden or assumed away. We argue that, based on our experimental study, it is these internal features and their interplay that will help provide the missing but significant insights to designing high-performance and high-endurance SSDs. In this paper, we use our highly accurate and multi-tiered SSD simulator, called SSDsim, to analyze several key internal SSD factors to characterize their performance impacts, interplay and parallelisms for the purpose of performance and endurance en-hancement of SSDs. From the results of our experiments, we found that: (1) larger pages tend to have significantly negative impact on SSD performance under many workloads; (2) different physical-page allocation schemes have different deployment en-vironments, where an optimal allocation scheme can be found for each workload; (3) although advanced commands provided by flash manufacturers can improve performance in some cases, they may jeopardize the SSD performance and endurance when used inappropriately; (4) since the parallelisms of SSD can be classified into four levels, namely, channel-level, chip-level, die-level and plane-level, the priority order of SSD parallelism, resulting from the strong interplay among physical-page allocation schemes and advanced commands, can have a very significant impact on SSD performance and endurance.