In the trace-driven approach, data traced on a real running system are used to drive the model; experimental model based on this approach is used, to study the effect on performance of various multiprogramming and multiprocessing system design choices. Model implementation is focused on those aspects of the system contributing substantially to total system performance.