The performance of non-preemptive scheduling policies in multiprogrammed parallel processing systems is investigated. A large body of literature in this field already exists. Through discrete event simulation, the authors provide some observations. For example, using execution rate characteristics such as average parallelism and processor working set in processor allocation does not improve system performance. However, using even a single preemption per job could reduce mean response time. Since the conclusions are drawn from simulations for some specific workload model, one must be careful in using the results.