Gracefully degradable disk arrays

The problem of designing fault-tolerant disk arrays that are not susceptible to 100% load increases on the functional disks when one of the disks in the system fails is addressed. A technique that combines the advantages of parity schemes and the traditional dual copy methods and offers a wide variety of options in providing fault-tolerance is proposed. A theoretical framework for solving the problem is presented and a number of constructive techniques are proposed. By utilizing the same amount of hardware as the earlier methods but with a better data organization and a different reconstruction technique, the system yields better performance during a failure. Merging two parity groups as a reconfiguration strategy is shown to have a number of benefits, such as reduced hardware overhead and improved reliability. A combination of block designs and the proposed reconfiguration strategy results in a highly reliable disk array with the same or less overhead as the earlier approaches and better performance during a failure.<>