The proposed research has as final objective advancing knowledge on disk arrays performance, as well as developing new disk array organizations and algorithms for improved performance and reliability. The following specific areas will be investigated. Mirored disks (RAID1) double the disk access bandwidth for reads, but even higher performance is possible through scheduling. A combination of single disk scheduling methods adapted to mirrored disks and methods utilizing non-volatile storage to defer writes will be investigated. RAID6 disk arrays tolerating two disk failures utilize two parity disks or one parity disk and one utilizing the Reed-Solomon code. Two data layouts for the former case have been proposed. The performance of these configurations will be compared with each other, as well as RAID5, while the system is operating in normal, degraded, and rebuild modes. Active disks process data locally to disk, responding with final or filtrated data rather than raw data. Techniques to process more sophisticated queries than those considered so far will be investigated. Prioritized processing of disk requests allows background applications to make progress without significant impact on the performance of higher priority applications. We plan to investigate nonpreemptive as well as preemptive priorities in processing disk requests.