Disk System Architectures for High Performance Computing

Because of the mismatch between I/O and CPU speeds, high-performance computers have long been forced to confront the fundamental I/O bottleneck. As processing power and memory size continue to grow rapidly for micro- and minicomputers, they too will become I/O limited. A number of hardware and software approaches such as parallel read-out disks, expanded storage (for example, solid-state disks), and disk striping, have been used to increase I/O bandwidth and thus narrow the CPU-I/O performance gap. In addition, new developments driven by advances in small-diameter (5.25-in and 3.5-in) disk drives, promise very high I/O bandwidth if large numbers of devices can be organized into arrays of disks. In this paper we review the state of the art in disk devices and I/O controllers, and will describe new approaches for very high-performance I/O based on redundant arrays of inexpensive disks (RAIDs).