Techniques to Quantify SCSI-2 Disk Subsystem Specifications for Multimedia?
Shahram Gandeharizadeh, James Stone, Roger Zimmermann
Department of Computer Science
University of Southern California
Los Angeles, California 90089
March 20, 1995
Magnetic disk technology has established itself as the mass storage device of choice in the commercial arena. Most often, the operating system of a hardware platform that employs this device (e.g., a UnixTM-based workstation) hides its physical attributes by conceptualizing it as an array of blocks. This abstraction expedites program development time because the programmer is no longer concerned with the working details of the mass storage device. Moreover, the final program is portable as long as a new target platform provides an identical abstraction of its magnetic disk drive.
This paradigm is effective for those applications with no real-time constraints. However, certain applications (e.g., multimedia) cannot tolerate significant variations in the service time of the disk drive. They must estimate the service time of the disk accurately in order to both support real-time constraints of the application and schedule the disk effectively in the presence of multiple requests. Assuming a Unixbased workstation, this study reports on our experimental techniques to identify the physical details of a magnetic disk drive. The models constructed based on the results obtained from these experiments can be used by a program to satisfy its real-time constraints.
The magnetic disk drive technology has benefited from more than two decades of research and development. It has evolved to provide a low latency (in the order of milliseconds) and a low cost per megabyte of storage (approximately 50 cents at the time of this writing). It has become common place with annual sales in excess of 30 billion dollars [oST94] (far exceeding that of both CD Drives and Tape recorders whose combined sales is less than 10 billion dollars). Both workstations and personal computers employ magnetic ?This research was supported in part by the National Science Foundation under grants IRI-9110522, IRI-9258362 (NYI award), and CDA-9216321, a grant from DOD/Intelligence Community, and a Hewlett-Packard unrestricted cash/equipment gift.