Hammerora

Thanks Steve.

We are testing with keyandthink set to false. But that is by intent - the reason being, when we tried with keyandthink set to true, it hardly stressed the CPU i.e. cpu usage was very low. Our main objective is to run 'x' load on both hardwares (the machines have ample cpu capacity).

By 1 thread are you referring to 1 vuser? If not, thread means the thread on server running the tool. And this thread is the process that generates the load simulating an application - correct? We have tried with 8 but could not really see any difference and so we have stuck to 1.

So that contention is reduced, we have 200 warehouses and then tried with 50 vusers but that hardly generated any cpu load on the server. Even with 750 vusers and 200 warehouses we dont see cpu utilization hitting any more than 8%. Ofcourse, IO is becoming a bottleneck at that point but that apart we see a huge amount of time spent in row contention. We also understand that if row contention is resolved it wont make much of a difference because IO is already a bottleneck and hence it will only severe.

BTW, can keyandthink time be changed? We did not see any such option in the GUI.

thanks.