I have an 8-core Xeon server.

 

Let's say I were to create a Linux VM and allocate it 6 cores, and also a Windows 10 VM with 4 cores.

 

Clearly they can't both be running at the same time ...

 

When a VM is shut down, are the cores released for use by other processes?

 

(I apologize if this is in the documentation, but I didn't find anything there)

Actually both VMs could be running at the same time. You can oversubscribe CPUs, but performance will be impacted as a result.

 

And to be clear, the way CPUs are assigned today, even when a VM is running, other host-based processes can utilize the same CPUs at the same time.

 

If you wish to reserve specific CPUs on your system for VMs only (and prevent all other host activity from hitting those cores), look up the isolcpus= Linux kernel command to do that.  We plan to implement that as an option for people to tune in the future.

Thanks for the response.

 

Although you didn't state it, your response implies that when the VM's aren't running, their CPUs are re-allocated to other processes.

 

Any plans for features like the POWER hypervisors have, like partial core (like 0.2), weighted allocation (i.e. VM 1 gets priority over VM 2), or desired vs. required allocation?

 

 

Thanks for the response.

 

Although you didn't state it, your response implies that when the VM's aren't running, their CPUs are re-allocated to other processes.

 

Any plans for features like the POWER hypervisors have, like partial core (like 0.2), weighted allocation (i.e. VM 1 gets priority over VM 2), or desired vs. required allocation?

 

Yes to your first question.

 

As far as the other features you mention, those would be a long ways out yet. We are not positioned as a data center solution for hosting enterprise workloads, for which those types of features would be better suited. For a home / SMB system, those features aren't nearly as important as cloning/snapshots, better usb/pci device assignnent, and others.