A trace-driven modeling techniques for computer system evalution is discussed. This approach intends to solve the inherent dilemma in computer systems modeling of too many simplifying assumptions or of too much detail. An experimental model based on this technique is also discussed. The objective of the model is to study the effect on performance of various. Model implementation is focused on those aspects of the systems contributing substantially to total system performance. While the operating system is conceptually modeled, detailed logic and timing are supplied in the job-trace profile, reducing modeling effort and improving model flexibility.