How reliable is a RAID?

Abstract

Disk arrays offer greatly increased transfer bandwidth at low cost, but without additional data redundancy they can suffer from significantly degraded reliability. The authors examine the reliability of RAID (redundant arrays of inexpensive disks) systems and find that although extremely reliable disk arrays cannot be attained with data redundancy alone, RAID system reliability can be made better than conventional large disks with little extra hardware. They show that the rest of the system components, such as the host bus adapters, power supplies, and fans cannot be ignored. The authors present a scheme in which the system support components are organized into groups orthogonal to the data redundancy groups, thus guaranteeing that no single disk or component failure will permanently lose data. This approach yields disk arrays with reliability about 50% greater than conventional disk subsystems. If further reliability improvements are sought, various parts of the support hardware can be made redundant. The authors have shown how these affect the design of a 49-data-disk RAID. >