DocumentCode :
1680408
Title :
Practical self-repairing codes for distributed storage
Author :
Hanxu Hou ; Hui Li
Author_Institution :
Shenzhen Key Lab. of Cloud Comput. Technol. & Applic., Peking Univ., Shenzhen, China
fYear :
2012
Firstpage :
76
Lastpage :
81
Abstract :
In distributed storage systems, projective geometric self-repairing codes (PSRC) are able to provide a repair mechanism, which can repair a failed node with only two nodes. However, the redundancy of PSRC is extremely large, which makes PSRC unpractical for distributed storage systems. In this paper, we propose a new family of self-repairing codes called practical projective geometric self-repairing codes (PPSRC), which is a generalization of PSRC. PPSRC allow tradeoff between the coding rate and self-repairing capacity, while possess the basic property of PSRC. In addition, we provide an evolutionary repairing process for PPSRC, of which the repairing bandwidth is less than both the repairing process in [5] and that of the minimal storage repair codes (MSR) [3]. Furthermore, we present the general reconstruction process of PPSRC and the reconstruction process is optimal in terms of downloading bandwidth. Finally, we demonstrate PPSRC surpass the existing erasure coding strategies in terms of complexity, repairing nodes, repairing bandwidth and self-repairing capacity, which makes the PPSRC very attractive for distributed storage systems.
Keywords :
codes; distributed processing; storage management; MSR; PSRC; coding rate; distributed storage systems; downloading bandwidth; erasure coding strategy; evolutionary repairing process; general reconstruction process; minimal storage repair codes; practical projective geometric self-repairing codes; repair mechanism; repairing bandwidth; repairing nodes; repairing process; self-repairing capacity; coding rate; complexity; distributed storage; self-repairing capacity; self-repairing codes;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Cloud Computing Congress (APCloudCC), 2012 IEEE Asia Pacific
Conference_Location :
Shenzhen
Print_ISBN :
978-1-4673-2903-3
Electronic_ISBN :
978-1-4673-2902-6
Type :
conf
DOI :
10.1109/APCloudCC.2012.6486516
Filename :
6486516
Link To Document :
بازگشت