Modern processor technology advances much faster than magnetic recording technology. This mismatch makes disk system performance a dominate factor in overall system behavior. There is need to construct simulation models of disk systems to determine and quantify the performance impact of alternative designs.
This thesis desribes a system-assisted disk I/O simulation technique. This technique uses the real system to help simulate, relieving the burden to model the target system. It accepts input trace data and returns estimates of disk request service times. In addition, we describe a disk trace colleciton technique capable of collecting accurate trace data from the Windows 95 operating system. This simulation technique is evaluated and found to be an accurate predictor of disk service times.