On the effect of a configuration choice on the performance of a mirrored storage system
Journal of Parallel and Distributed Computing
| Hi-index | 0.00 |
Parallel disks can improve I/O performance in a manner analogous to the use of parallel processors to improve computation times. However, due to their data storage function, reliability issues become exceedingly important. Currently proposed schemes use shadowing or parity to achieve reliability in parallel disks. In this paper we introduce the idea of using the Information Dispersal Algorithm (IDA) of Michael O. Rabin [8] to distribute data and redundancy information uniformly among multiple disks and compare the performance and reliability characteristics of shadowing, parity, and IDA. We discuss some ways to take advantage of the uniformity of data placement and argue that IDA is the algorithm of choice for achieving reliability and performance in parallel disk systems.