I've just installed NetBench and have done a few runs against a NetApp server with JBOD and two 733MHz PIII clients all on gigabit ethernet. The through ut is not that much different than when the clients were on 100BaseT.

My question is: Does the standard NetBench disk mix suite have a bunch of sleeps or something that would cause this? I notice that the published results views have throughput on the y-axis and number of clients on the x-axis. This would seem to suggest that they have, through coding, genericized clients so that fast clients aren't a penalty. Is my guess right? If so, is there a way to disable the sleeps so that this benchmark tool could be used as a stress tool?