We are in the process of receiving two sets of test hardware to run a VMware ESX 3.5 cluster with the full blown HA, Vmotion, and extras. We would like to compare these hardware sets by building as many copies of a template virtual machine as we can until either the software or hardware begins to fail (can I say I love my job?). In any case, I need some opinions on how to go about testing a virtual machine. One thought I had was to boot into prime95 or similar software and run the same stress tests concurrently on all the VM's as I am adding them to the cluster for consistancy. I was also thinking of building a template Centos machine and running a stress test from within the OS that maxed CPU, Memory, and Network concurrently (haven't found anything to do this yet, suggestions welcome).
We are getting some of the new nehalem stuff to test, and we would like to justify the extra cost and risk of new hardware by testing the heat-sinks out it and proving it will hold up and spank the older Core2/Opteron architectures.
Thanks in advance!