Thanks for the interest Ken.
the equipment is certainly dated, the msa1000 was not intended for this environment and the customer is cost justifying a c3000 with 4 full height 4p quad core and a new msa2000 (2300??).
i have always had a concern about the msa1000 throughput but in this case it does not appear to be overloaded and moving a vm onto the local disk of a server with no other vm's present is as light a load on a server an i/o as is possible (not that the local disk is very fast) considering that the guest xp workstation has booted and done nothing else so it is idle and not using much more than 1-3% cpu and not doing i/o after the boot is complete. Yet the CPU Wait time is still around 18,000 milliseconds on average.
i tried esxtop and the %ready is low to none because the guest is not doing anything. on the other esx hosts with guests actually doing stuff the reading for %ready is quite low typically and consistent with the CPU Ready time reported in the VC performance graphs.
so in general i am thinking the CPU Wait time figure can be a bit of a red herring in that on an idle system it is very high and on an active system it drop down consistent with the actual cpu load going up. i do not know what that perf counter was intended to be but at this time it does not appear to be reporting i/o blocking, at least in this case.
i intent to get support involved but would first like to determine a likely area of investigation, at this time there is nothing in particular that jumps up and waves a red flag, yet the performance of everything is slow, in particular the part about the windows task manager indicating 100% cpu utilization yet esx logs what appears to be rather normal cpu utilization certainly nowhere near the 100% windows thinks it is using.
thanks
