Performance differences and higher cpu usage
-
Subject: Seeking Advice on VM Performance Issues Post Migration to XCP-ng
Hi everyone,
As we wrap up our migration from VMware to XCP-ng, we've encountered some challenges. Notably, a few customers have reported that their VMs are running slower than they did on VMware. While I haven't noticed significant differences in the performance of my own VMs, some do appear to be underperforming. We primarily host Windows Server VMs with Remote Desktop.
We've observed notably higher CPU usage in Chrome and Outlook. This might be related to hardware rendering or how images are displayed. Currently, the VGA switch is off on all our VMs—a default setting applied by XCP-ng during the migration. I'm curious if turning this switch on would improve performance, especially for Remote Desktop usage, although support suggests it would.
Has anyone dealt with similar issues or settings adjustments in Windows VMs with Remote Desktop? Any insights or real-world experiences would be greatly appreciated.
Thank you!
-
Hi!
How did you migrated those VMs in the first place?
-
@olivierlambert good question, we used the builtin feature of XOA.
-
Okay, so first thing you can try:
- Shutdown the VM
- Create a new VM from scratch using the "right" template that matches the OS used on the original VM (ie if it's a Windows 2019, use a Windows 2019 template and so on)
- Don't install anything, don't boot it, just create it
- Attach disks from your previously shutdown VM
- Start the new VM with the freshly attached disk
You might have already better performances in theory. Please let me know if it's the case.
-
@olivierlambert ok, will try this on one of the vms as well. what about the VGA switch? does this affect graphics inside the vm?
-
On Windows I'm not 100% sure of the impact. I would try first with the new template, and if it doesn't change anything, use the VGA stuff.
-
@olivierlambert We observed that a VM with higher memory usage on XenServer reacts more slowly compared to VMware. This might be related to the different functioning of XenServer tools. We resized the VMs where customers had complained, and this led to significant improvements.
Additionally, we adjusted the VGA switch. While it's unclear if this had a major impact, after increasing the memory and setting the switch, we received positive feedback from the previously dissatisfied customers.
