Hello,
I have two Hyper-V Hosts, one on Windows Server 2016 and one on Windows Server 2012 R2.
The host machines both have 64GB of memory installed, 20 guests with a startup memory of 2048MB (no dynamic memory), and 10-11 free GB of memory (in task manager).
The guests are running Windows Server 2012 R2.
Every guest in this scenario had a checkpoint made programmatically after the initial boot and some setup of our business logic.
Restoring checkpoints and waiting for the VM heartbeat is much slower in 2016.
Example: 2012 R2 takes 3-4 seconds, 2016 takes 12-20 seconds.
Also, the Memory/Pages/sec performance counter can easily reach into the thousands on the 2016 box.
However, the Process/Pages/sec performance counter (_Total) is much lower, indicating that the page faults are from kernel-mode processes.
When setting the configuration version for the VMs to be 5.0 on the 2016 box (2012 defaults to this, 2016 defaults to 8.0) the restore times did go down, but not to the levels of the 2012 R2 host.
Are there any configuration values I am missing here that might help me?
Are there any performance counters that might be helpful at pinning the root cause?
Note that the checkpoints are standard checkpoints, not production.