Trade-offs between devices and paths in achieving disk interleaving

• Authors: S. Ng;D. Lang;R. Selinger
• Venue: ISCA '88 Proceedings of the 15th Annual International Symposium on Computer architecture
• Year: 1988

There is a continuing need to improve the performance of disk subsystems, and one of the key factors of a disk subsystem's performance is the data transfer rate. While it is clear that increasing the data transfer rate would reduce the service time for those applications that move large amounts of data per I/O, the amount of data transferred per I/O in many of today's typical commercial applications is relatively small. The benefit of a higher data rate in these environments was not well understood, but is explored here and is found to be larger than expected.One way to increase the data rate of the disk subsystem is through a technique generally known as striping. Using this approach, multiple conventional disks are grouped together and function as if they were a single one, with the data spread among the multiple disks and transferred in parallel. In this paper, we study four alternative implementations for achieving higher data rates in a disk subsystem, focusing on the trade-offs between the number of devices and the number of data paths, keeping the number of physical devices constant (which may keep the cost roughly constant). The performance advantages and limitations of the alternative implementations are analyzed using an analytic queuing model and compared to a conventional disk subsystem.