I am having lots of performance issues after upgrading everything to vSphere 5 and View 5 when connecting using PCoIP. The interaction with the VM's are slower with either dual or single monitor. I have tried all of the different workarounds suggested from the Teradici site. If anyone has any fixes or suggestions to how to clear things up please reply.
add the following line to the vm config file.
mks.poll.headlessRates = 1000 100 2
It should fix your PCoIP lag issues, let me know what your results are.
Hi Darnell,
Any chance of getting some screenshots of your performance on disk IO/ network and CPU/memory?
Upgrading to View 5 should actually net you better performance, so this is a bit puzzling.
Cheers!
We noticed that same PCoIP lag issues after we upgraded to View 5 from View 4.6.
I called VMware support and they gave us a entry to add to the config file of the VM to fix the issue.
Once we added the entry and rebooted the VM the PCoIP lag issue was fixed.
What is the entry that VMware provided?
I am not at work at the moment but I will login and get it for you.
Stand by.
Ok. Thanks...
I'm curious too. I used 4.6 a year ago to test out for a month. Then it wasn't until 5 came out that I touched it again and started to do a production deployment. It's been a long time since then, but I swear 4.6 seemed smoother. I'm very curious about this entry too.
add the following line to the vm config file.
mks.poll.headlessRates = 1000 100 2
It should fix your PCoIP lag issues, let me know what your results are.
Awsome!!! This setting did the trick. Thank you.
I've had the same performance issues and after adding this Parameter all issues are gone.
After a lot oft searche requests on google and the VM Knowlagebase I still don't know what this parameter exactly do.
Is anybody able to explane this parameter?
Well at least we know the MKS stands for mouse, keyboard, and screen. I'm not sure what the rest stands for.
Just add this to the Advanced Configuration Under "Options, Advanced, General, Add New Row"?
I asked my VMware Federal guys what this setting does and here is the answer:
mks.poll.headlessRates sets the frequency that the ESX host processes graphics commands from the SVGA FIFO.
ESX slows the MKS poll rates when there are no VC consoles attached. It's an optimization that reduces MKS overhead on server workloads. (slower poll, less CPU). Normally this does not affect View because the Graphics team has been working for a while to make the MKS event driven, reducing the work that happens in the poll loop. Therefore, we don't expect this setting to affect View performance. However, we found a bug where the act of moving a window from one display to another depends on some work done in the poll loop. Therefore, when a VM has no VC consoles attached, and a View user drags a window across monitors, graphics processing performance is limited by the server-optimized poll rate. This leads to the "extreme jitter" bug.
The workaround is to add the headlessRates .vmx entry, causing the "slow" server-optimized poll rate to be as high as the "fast" console-attached poll rate. Essentially removing the server/idle workload optimization.
The only negative consequence for adding this .vmx option is potentially increased idle CPU usage, however I don't think anyone here can quantify how significant it would be. Engineering does not see any way this could affect security.
I added this setting to my View pools and noticed a slight change with Aero disabled. With Aero enabled, it was more dramatic but still not as fast as View 4.6 was.
Great information Cannoli. Thanks for sharing.