Hi All,
Its a Monday, and this Monday hasn't started well! My brain is trying to work out the answer to a question and I'm struggling to work it out in my head so opening the floor up here for someone to give me a decent answer (any put my brain out of its misery!).
Ok this is a theoretical question based upon expected behaviour from the guest OS with differing pCPU sizes on vSphere hosts within the same cluster.
I have one vSphere cluster with two differing pCPU configurations. The CPU instruction sets are the same but the clock speed on the processor is different. One processor is a 2.0Ghz clock speed and the other is a 3.0Ghz clock speed. What happens with DRS in this scenario and also what would happen if a highly utilised VM was vMotioned from one host to another?
For example, if a VM which is running on the 3.0Ghz, is consuming 80% of its allocated vCPU and it then gets vMotioned from that host to one of the hosts with a smaller pCPU clock speed, what would happen? As the VM cannot have more pCPU than the underlying hardware, would the VM hit a CPU bottleneck?
I know that its best practice to try to keep specifications for vSphere in a cluster similar where possible, and that EVC will help limit changes in CPU instruction sets when enabled but my question is not around this - its regarding the effect of differing pCPU clock speeds within a cluster and the effect on underlying resources provided to the VM's Guest OS in the even of migration. As hardware specifications, move forward and specifications change, obviously its difficult to keep the same specification and mixed hosts are almost an inevitability unless additional cost can be tolerated to provide N+1 for every new blade spec. Unfortunately, this is not always possible from a commercial perspective.
I hope I'm making sense! Thanks in advance.
Cheers,
Warren