Working on providing provisioining for a new VM host and I came across this:
The first reply mentions that vsphere will provide out of order execution (kind of running stuff in parallel in hypervisor?) for a single vCPU to allow it to perform faster than a single pCPU would. Is this actually true? I have not found anything in the documentation (the whitepapers provided on the cpu scheduling) that supports this claim. As far as I can tell, a single vCPU will only ever execute it's unit of work on one pCPU at a time (although not necessarily the same pCPU every time). That is to say if a vCPU was 100% utilized while executing a single thread for a host OS, the underlying vsphere would only see 100% of one pCPU used at any time and we would never see 100% of two pCPU's utilized (effectively 200% utilization).