(2011-2012) Parity Declustering for RAID systems

RAID (Redundant Array of Independent Disks) was developed in 1980s to provide large-scaled, highly reliable data storage systems. A problem with RAID is that as the disk capacity increases, the reconstruction process when disk failure occurs becomes much slower. One of our main goals is to design an erasure-correction code coupled with a data layout scheme on disk arrays that maintain the high reliability of the current system and at the same time have shorter reconstruction time.

Parity declustering is a well-known technique that allows faster reconstruction of a disk array when some disk fails. Moreover, it guarantees uniform reconstruction workload on all surviving disks. It has been shown that parity declustering for one-failure tolerant array codes can be obtained via Balanced Incomplete Block Designs. Our major contribution in this line of research is to extend this technique for array codes that can tolerate an arbitrary number of disk failures via t-designs, a well-studied object in combinatorial design theory. Our work is summarized in the following manuscript:

S. H. Dau, Y. Jia, C. Jin, W. Xi, and K. S. Chan, “Parity Declustering for Fault-Tolerant Storage Systems via t-designs,” 2012.