CPU Ready time question (PCPU vs VCPU)

[grantcunningham]

Hi All,

I'm trying to validate CPU Ready % for some of our clusters and was wondering why there is a difference in PCPU Ready % vs VCPU Ready %

I'm looking at PCPU Ready % on a ESX host and I can see the Summation value of 7000MS in the real time charts. When converted this gives me a percentage value of 35% (CPU summation value / (<chart default update interval in seconds> * 1000)) * 100 = CPU ready %)

Am I calculating this wrong?  This seems extremely high.

When I look at ESXTOP on that same host, I can see the VCPU's Ready % (%RDY and also looking at %VMWAIT?) all well under vmware's recommended best practice of %5

My question's are;

1. Am I calculating this value correctly?

2. Why do my VCPU's show a different percentage than my PCPU? I would of thought that they would be aligned.

3. Should I calculate my CPU Ready % on a per PCPU basis (eg. %RDY or %VMWAIT)

4. Why is my PCPU Ready time so high if the VM's CPU Ready % are so low.

Any help would be much appreciated.

Thanks.

[vNEX]

Hi,

You forget to divide it by number of logical CPUs in your host or by number of vCPU in your VM!

In ESXTOP you see all these metrics in real-time view which is refreshed every 5 seconds ... there is no need for such formulas above.

The value you see in ESXTOP for a VM is aggregate sum of %RDY for all vCPUs dedicated to VM so for SMP virtual machine you must divide this number to number of VM vCPUs.

If you wan to see %RDY for every vCPU separately in ESXTOP push "e" to expand VM worlds and type VM GID in the list you will see %RDY for each vCPU (vmx-vcpu-0, vmx-vcpu-1, vmx-vcpu-2 ....) separately.

I'm looking at PCPU Ready % on a ESX host and I can see the Summation value of 7000MS in the real time charts. When converted this gives me a percentage value of 35% (CPU summation value / (<chart default update interval in seconds> * 1000)) * 100 = CPU ready %)

Am I calculating this wrong?  This seems extremely high. Am I calculating this value correctly?

So for dual socket quad core host with HT enabled the calculation would be following:

7000/20000*100/16= 2,1875 % of ready time per logical CPU which is absolutely no problem at all.

%VMWAIT - means time VM spent in blocked state waiting for some events to complete; usually waiting for I/O

High VMWAIT can be caused by poor storage performance or high latency of some pass-through device configured for the VM.

_________________________________________________________________________________________

If you found this or any other answer helpful, please consider to award points. (use Correct or Helpful buttons)

Regards,

P.

Message was edited by: vNEX