DISK ARRAYS - HIGH-PERFORMANCE, HIGH-RELIABILITY STORAGE SUBSYSTEMS

Storage subsystem performance and reliability can be enhanced by logically grouping multiple disk drives into disk arrays. However, many combinations of data distribution schemes and redundancy mechanisms are possible in an array data organization, so trade-offs between performance and reliability must be considered. The authors explain how disk striping is used to distribute data in the array and how the choice of stripe unit size affects performance- They also discuss data replication and parity-based protection, focusing on their implementation in disk arrays as a way to deal with disk failure. Should a failure occur, lost data can be reconstructed from the surviving disks. Of course, performance is usually affected while this rebuild is taking place, and an additional failure at this time could be critical. Striking the right balance between performance and reliability is an important consideration for users, and techniques are available for improving both. These include the use of nonvolatile RAM, write remapping, and logging. The promise of additional developments in disk drives suggests that disk arrays will become the implementation mechanism of choice for secondary storage subsystems in the near future.