SGER: Designs for High-Performance Fault-Tolerant Disk Arrays

This is a planning award to investigate systems and software aspects of RAID systems in which system development can benefit from theoretical research. The goal is to formulate research problems that combine theory and systems efforts to affect the design and performance of the next generation of RAID systems. The focus will be in four areas: data layouts that tolerate multiple faults; on-line extendibility of disk arrays, in which new disks are added to a running system; efficient implementation of data-layout mappings; and the effect of system-level optimizations such as large read and write optimizations. Collaborations with experimental RAID groups (such as Carnegie-Mellon University, UCSD, and IBM Almaden) will be pursued, with the goal of obtaining new research topics and new metrics.