A correlation-aware prefetching strategy for object-based file system

  • Authors:
  • Julei Sui;Jiancong Tong;Gang Wang;Xiaoguang Liu

  • Affiliations:
  • Nankai-Baidu Joint Lab, College of Information Technical Science, Nankai University, Tianjin, China;Nankai-Baidu Joint Lab, College of Information Technical Science, Nankai University, Tianjin, China;Nankai-Baidu Joint Lab, College of Information Technical Science, Nankai University, Tianjin, China;Nankai-Baidu Joint Lab, College of Information Technical Science, Nankai University, Tianjin, China

  • Venue:
  • ICA3PP'10 Proceedings of the 10th international conference on Algorithms and Architectures for Parallel Processing - Volume Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

The prefetching strategies used in modern distributed storage systems generally are based on temporal and/or spatial locality of requests Due to the special properties of object-based storage systems, however, the traditional tactics are almost incompetent for the job This paper presents a new prefetching approach, which takes the correlationship among objects into account Two orthogonal replica distribution algorithms are proposed to aggregate prefetching operations A moving window mechanism is also developed to control prefetching We implement these approaches in our object-based file system called NBJLOFS (abbreviated for Nankai-Baidu Joint Lab Object-based File System) The experimental results show that these approaches improves throughput by up to 80%.