1. The Host – Total CPU Usage is completely different from what is reported in Task Manager for a virtual windows machine? Therefore, the two values cannot be compared as one is for ESX server(Host) and the other for the virtual machine.
2. 'CPU Virtualization Overhead shows if ESX is restricting the virtual machine's access to the CPU' according to the Help file. What would a 100% overhead signify?
3. Is virtualization overhead value expected to go up or down with changes in the Host – Total CPU Usage value? How does these values affect each other.
4. How is the value of virtualisation overhead calculated?
1. Correct. For monitoring Windows, we subtract an Average VM CPU value from the reported Average CPU usage from the WMI metrics. This gives, we believe, a true CPU usage for the VM. When monitoring Unix the metrics actually return the adjusted value. (These metrics are reported by the VM image itself from the OS and not from ESX metrics).
2. That the VM is being allocated and is using its full amount of CPU allowed - defined by a combination of ESX limits, capacity, reservations and VM usage.
3. Yes. See 4 below.
4. ESX provides a lot of metrics on the CPU usage etc for a VM which we use to work out how the CPU performance of the VM is being affected.
The calculation goes sthus:
i) If the CPU has been limited for the sample period calculate the "Capped CPU %" (from CPU Limits, CPU Capacity and CPU reservation)
ii) Calculate the amount of "Ready CPU" provided by ESX
a. from Ready CPU % or
b. scaled Ready CPU % according to the total CPU capacity
iii) Overhead CPU = smaller value of the Capped CPU or Ready CPU, plus ESX reported VM system CPU
So as the ESX restricts or caps the amount of CPU "given" to the VM the overhead rises, it also rises as that amount of CPU is used.
Our goal is to show if your VM is being allocated enough resources (CPU in this case) by the ESX.