Hi,
I came across the following point in a CPU Best practice document.
"Having virtual machines configured with VCPUs that are not used still imposes resource requirements on the ESX server. In some guest operating systems, the unused virtual CPU still consumes timer inerrups and executes the idle loop of the guest OS, which translates into real CPU consumption."
Can someone provide some input on the bolded section.
Why is it that if the idle loop executes timer interrupts and the idle loop of the guest OS will cause real CPU consumption?
What example OS can do this?
Is it true that "generally" an app that does not understand SMP will run slower on a multiple CPU hardware?
Hi,
. In some guest operating systems, the unused virtual CPU still consumes timer inerrups and executes the idle loop of the guest OS, which translates into real CPU consumption."
Running Virtual machine is a process for a hypervisor and VMM(Virtual machine monitor) will monitor he VM processes and execute the request given by the VM. Each VM is having overhead or penalty of their resources which has to be executed on pCPU.
Can you tell us how you are measuring real CPU consumption?
Is it true that "generally" an app that does not understand SMP will run slower on a multiple CPU hardware?
If your Application dosen’t understand the SMP and you are having multiple vCPUs on that VM then its simply wastage of resources. it can slow down performance by increasing RADY time for requested processes by other VMs.
If your virtual machine is not running multi-threaded application you might see a high %cstop value in esxtop. For more information about %cstop read this article:
I've written a couple of articles about rightsizing the virtual machine, maybe it can help you sizing your virtual machines:
http://frankdenneman.nl/memory/impact-of-oversized-virtual-machines-part-3/
http://frankdenneman.nl/vmware/impact-of-oversized-virtual-machines-part-2/
http://frankdenneman.nl/vmware/impact-of-oversized-virtual-machines-part-1/
http://frankdenneman.nl/memory/sizing-vms-and-numa-nodes/