Architecture of a fault-tolerant RAID-5+I/O subsystem

An approach to the design of a fault-tolerant input/output subsystem is emphasized that focuses on basic architectural and systematic requirements. The MTI FailSafe product is a fault-tolerant RAID-5+system. FailSafe supports a high mean time to loss of data availability (MTDA) through fully redundant hardware, a verified data path, and a design philosophy that, while not attempting to be verifiable, accepts and prepares for errors with an architected recovery policy. RAID provides the no-single-point-of-failure property for the media and, with a hot spare, can provide a fairly seamless repair cycle for failed drives and drive channels. The author discusses what he believes a fault-tolerant RAID system architecture beyond RAID should consist of. The FailSafe system architecture is presented. The key implementation and design tradeoffs are considered. Some of the performance issues and operational requirements for RAID systems are discussed.<>