High guest OS resource utilization

Performance metrics within a Virtual Machine's Guest OS are different (higher or lower) than the same metrics reported by Veeam ONE.
 

The process used by vSphere to allocate and distribute system resources differs from the methods used by Windows and other operating systems for a physical machine. As Veeam ONE collects its performance data directly from vSphere, the metrics you see in Veeam ONE reflect those found in vSphere rather than VM’s guest OS. The guest OS can only view and analyze performance from within the virtual machine, whereas vSphere uses an overall view of all virtual machines and resources to calculate and analyze performance.
 
There are several reasons why the metric values may differ:

1. Virtual machine’s guest OS is unaware of hypervisor resource overheads. This occurs because the guest OS is only permitted to view a certain portion of all available resources as determined by hypervisor’s scheduler.

2. Resource overheads often change based on workloads.

3. There may be a time difference between performance data displayed at the guest and hypervisor level. Additionally, the time intervals at which performance data is recorded and displayed may differ between the guest OS’s resource monitor and the hypervisor’s resource monitor.

4. Because a virtual machine’s guest OS is unaware of its hypervisor’s resource distributions, it is therefore unaware of being “de-scheduled” while idle.
    

As a general rule of thumb, the performance metrics reported by Veeam ONE, and in turn vSphere, should be taken into consideration rather than guest OS’s performance metrics. If there is any doubt of the accuracy of Veeam ONE’s performance data, verify first that the data matches the corresponding data in vSphere.