As the VMWare wonâ t use the hardware clock, will this affect somehow the way LoadRunner manage all the timing values (the controller and the load generators.)
The VMWare will add errors to precision of measurements (e.g. transaction response time), however, the ESX server can be used and it depends what kind of a Load Test planned.
The precision will be to the millisecond, which in most cases is not so critical, since load test messuring seconds is enough (rather than milliseconds). On top of that. LoadRunner depends on hardware clock which is emulated in ESX.
If you want more details, you can get it from HP support knowledge base.
The trend toward VMs is, in some ways unsettling as the VM companies would like you to believe that you can get 2x performance out of a machine with 2VMs running.
if you are lucky and you are running an application that does not in any way take advantage of multi-core or multi-CPU machines you -might- see a boost, but you won't be able to get twice the performance out of it.
we regularly run out of cpu cycles well before we run out of memory in our testing here, and since we usually run each user as a process, Win2003 seems to do a pretty good job of splitting the processes across the 2 multi-core CPUs and the individual processor cores(supposedly equiv. to a 8-way server, but not quite)
recently i was asked why we don't just run 2 vm on our generator machine and the manager asking could not understand that running 2VM would not give us the capabilities that running 2 machines did.