Pool level resource settings vs VM resouce settings level

[gelico]

Hello,

In order to fix the cpu speed of a VM (RH 4 64 bits) in a narrow range

to avoid sync time problems, I configured a Resource

Pool with a Reservation and a Limit s values to 3000 and 3200 respectively.The Pool has just a VM.

My doubt is if I need to modify this setting at VM level (editing settings) too or is not necessary.

I say this because the CPU speed reported on virtual center to this VM is not the configured on the pool. Is this a normal behavior?

Is this the right way to fix the VM's cpu speed? Should I see the CPU speed is the one I set at resource pool level or VM level?

We are working with ESX 3.0.2 and VC 2.0.1

Thanks

[weinstein5]

If I understand your question you are looking to set the VM to a fixed number of Hz at times - really there is no way to do this. Remember, the speed of a processor is the number of instructions it can do per second - this defeats one of the benefits of virtualization of being dynamically map resource when needed so if your RH VM is not using those cycles then they will be made available to other VMs so with that in mind the best you can do is set a reservation for your VM of the number cycles in MHz - 3000 - this means when the VMs are fighting fro CPU cycles the VM will always have 3000 MHz if it needs it - and if it is not using the 3000 those other cycles will be given to other VMs - and if you want to cap the consumption your will set a limit for the VM of 3200 -

now why not do this at the resource pool level - the resource pool is used to is to create pools of resources to be used by VMs - which does not mean the VM will get that -

Now why doesn't VC show the speed in that range that is because VC shows the number of cycles your VM is actually using -

Hope this helps

[mcowger]

The right way to fix your problem is not with CPU reservations (which probably wont fix it anyways), but with NTP.

--Matt

[gelico]

Hello,

Yes, we are playing with ntp, vmware tools too, but the problem is much more complex.

The VMs are RH 4 64 bits in a cluster and each node are in different ESX host with different physical CPU (number and frequency)

Maybe I'll post later.

Thanks