¡¡Chinese Journal of Computers   Full Text
  TitleA Data Placement Based on Toleration on Triple Failures Array Codes in RAID
  AuthorsWAN Wu-Nan1) WU Zhen1) CHEN Yun1) WANG Xiao-Jing2)
  Address1)(Department of Networks Engineering, Chengdu University of Information Technology, Chengdu 610225)
2)(Chengdu Institute of Computer Applications, Chinese Academy of Sciences, Chengdu 610041)
  Year2007
  IssueNo.10(1721¡ª1730)
  Abstract &
  Background 		Abstract A novel data placement based on EEOD code in RAID has been presented. The EEOD is an extension of the double-erasure-correcting EVENODD, which can tolerate simultaneous failures of three member disks with only three extra disks for parity information. A rigorous theoretical proof of MDS property based on algebraic representation is given. The EEOD code¡¯s decoding algorithms is discussed by representing a check group consisting of date and parity units with graphics circuits. The decoding process can be regarded as superimposition of graphics circuits and vertices of even degrees in the graph are eliminated gradually. The encoding and decoding procedure of EEOD code is based on XOR operations, and the decoding complexity of the EEOD code is much lower than those of the existing comparable codes, while decreases the influences on system¡® s throughput and disk capacity. Thus the EEOD code is of practical signficative for storage systems.

keywords EVENODD; RAID architectures; array codes; data placement; MDS

background This research is supported by the National Basic Research Program(973 Program) of China under grant No.2004CB318003 and SiChuan Science and Technology Tackle Key Problem Program under grant No.03GG0326 and KYTZ200706. As with computing, fault tolerance (or reliability) is increasingly important in storage systems. Some critical data should be available and some services should be provided even when faults occur in storage units. The main purpose of the research is to study to improve high availability and high reliability of large scale storage systems, especially on Disaster Tolerance and Fault Tolerance of storage systems based on RAID technology. To improve the availability of user data, the authors propose a class of new binary Maximum Distance Separable (MDS) array codes called EEOD-Code. EEOD-Code is 3-erasure-correcting code which can provide a much longer nonstop functioning time to a distributed storage system.