In this paper we analyse three general classes of scheduling policies under a workload typical of large-scale scientific computing. These policies differ in the manner in which processoer are partitioned among the jobs as well as the way in which jobs are prioritized fro execution on the partitions. Our results indicate that existing static schemes do not perform well under varying workloads. Adaptive policies tend to make better scheduling decisons, but heir ability to adjust to workload changes is limited. Dynamic partitioning poicies, on the other hand, yield the best performance and ca be tuned to provide desired performance diffrences among jobs with varying resource demands