https://worldofeuc.com logo
Join Slack
Powered by
This message was deleted.
# citrix-vad
s
This message was deleted.
j
the first thing I'd check is that the vms are licensed properly and checking in with your license server CLS or legacy
j
Everything looks good from the licensing standpoint!
j
can you share your Citrix policy?
j
I should also mention these are MCS random machines and single session.
I know the host to client setting is really for server based multi-sessions, but I was giving it a shot to see if any difference. I can’t say it has had any impact…
p
Was the grid driver installed prior to the VDA?
j
I don’t believe so. I think I had the VDA installed and then completed the GRID install.
r
A few questions remarks: • Which vGPU driver version are you using? • What vGPU license are you using? • How many monitors / which resolution are you using • I would recommend to always install the vGPU driver before the VDA • Please remove the Desktop Composition Redirection and Queuing and Tossing policy. You don't need those • I would really recommend moving to VDA 2203, there are many GPU optimizations between those 2 releases especially since VDA 2112. Furthermore, what's the workload? When do you see the issue?
j
I am using the latest driver, maybe one behind since I think they just released a new version. I am using the 2B licensing for vPC. Two monitors primarily but would like to run up to 4. I will reinstall the VDA after the driver. Policies have been removed. I am currently on the LTSR track with 1912, and I am awaiting the release of the first CU on the latest LTSR before deploying.
r
Alright, I would recommend then at least to test with 2203 VDA. We (I work for Citrix Engineering) really changed the game with 2112 VDA: https://www.citrix.com/blogs/2021/12/16/hitting-new-highs-with-hdx-3d-pro/ and it's not only HDX 3D Pro that benefits from these improvements. On another note, if you are doing more than 2 monitors with a 2B profile I would recommend to look at your GPU memory usage. If that is constantly above 70% of the 2 GB you can run into performance issues as well.
j
Thanks! I will give this a look! Do you have any suggestions for web video, i.e. youtube or vimeo types sites when trying to get them to play and not be so choppy and distorted when playing inside of a VDI session for a user logged in from remote?
r
Well, the T4 should be able to decode that (YouTube uses VP9) and in Selective mode this is encoded with H.264 which is hardware encoded on the T4 as well. So if your endpoint can properly handle H.264 it should give a good user experience. You can see this on the VDA by looking at the Performance tab of the Task Manager and selecting the GPU. You would see Decoding & Encoding utilization. What endpoints are you using? Again, if it's Windows I would recommend CWA 22.06 which contains many performance optimizations as well.
j
I set the codec to full screen and optimized for 3d load to enabled and I am seeing amazing results. However, the gateway is also reporting a higher amount of bandwidth being transferred at the same time. I only have a VPX10, so I can see this getting saturated quickly. Are there any other suggested settings to alter to save on bandwidth and compress the video more? I am testing now by removing the optimized for 3d workloads to see if any better.
Mixed bag of endpoints. iPads/Chromebooks/Macbooks/Windows PCs
I have IGEL clients local on the network
r
What is your workload? It really depends on that which policies work best. VPX10 can definitely be a limiting factor here. How many sessions do you want to run through it?
j
What is the worst thing that will happen if the VPX hits its limits? Dropped frames or reduced quality until more network is available?
5-6 sessions
general task working but with web conferencing and other online training as well.
j
one other thing I would mention is enabling h.265 on the local windows endpoints. Its not available for non-windows
its pretty amazing when its enabled
r
Then our default policies are the best. So: Use video codec: use when preferred Quality: medium Use hw encoding: true @Josh Farlow Yes H.265 can help, but in this case only for videos. Also as you state, the endpoint needs to support it
j
most users are on apple devices, but I will keep this in mind for my few windows users! Thanks!
r
It depends on the endpoint if we use H.265 or not so there is no VDA policy you need to set for it. Only requirement at VDA side is a gpu with H.265 encoding, which is fine with the T4
j
well, I turned off the optimized for 3d loads and put back to preferred and the troublesome performance is back. I tried to turn it back and it certainly makes a difference, but would the overall machine performance be degraded or just the video and images? I am getting ready to reinstall the VDA and republish now.
p
Any events in event log?
j
nothing that I can tell
j
Hey @Rody Kossen - it's awesome having you here - just saying 🤘
💯 5
r
Thanks @James Kindon 😊
@Jeff Kater that behavior is definitely unusual. Let me know if it works better after reinstall
Could you share a video of the issue?
j
I was able to reinstall, but performance issues remain. I will see about grabbing a video once more network activity is present. Turning everything up to high worked better, but it also was using 6-7mb of my VPX10. It is acting as though it is struggling to identify the preferred video areas from static data. Repeating actions is a little better, I.e. launching file explorer. If you attempt other actions in between, lag is back.
r
Can you download RDanalyzer and share a screenshot of the test? Do you also see it when you connect directly and not via VPX? What is your endpoint / workspace app version? Would you be able to test with VDA 2203?
j
This was taken while connected via VPX. I will grab another for comparison when connected locally on the LAN. I am currently on a MacBook Pro using 22.06.1.51 (2206) to make my connection. I was able to finally get the analyzer to pick up on a few hardware encoding sessions, not very consistent in Chrome. The activity seemed to pick up more with Edge. When trying to do multiple things at once, i.e. Edge, MS Word, and then trying to launch file explorer, the lag is definitely there.
I did notice larger than normal memory usage on one of my delivery controllers. I was able to apply some updates and reboot. This may have made things slightly better?
r
This seems normal, from this view I don't see why there is any slowness. Also, the Delivery Controller is not in the ICA traffic path so it does not make any difference if you reboot it or not
j
I am going to try to build a new image with nothing installed in it and see if any better. I have also opened a case last week with Citrix for further troubleshooting, but I am still awaiting my initial contact from them.
r
Ok, you are using EDT here, did you configure EDT MTU Discovery? This is still needed on 1912 LTSR, with 2203 LTSR this is enabled by default: https://docs.citrix.com/en-us/citrix-virtual-apps-desktops/1912-ltsr/technical-overview/hdx/adaptive-transport.html#edt-mtu-discovery You could also try to force it to TCP
If your EDT MTU Size is off, you get also these performance issues
j
@Rody Kossen Thanks! I am looking into GPO settings now, as my new image I created seems much better in another OU. When I move to directory of production desktops, similar issues return. My old image is BIOS boot and new image issue UEFI. Otherwise all installs are the same.
r
@Jeff Kater Ah, yeah I would definitely investigate your GPO settings if you notice such a big difference.
j
@Rody Kossen While I am looking into these settings, just curious on the VPX10. When I am only testing one session through the VPX10 and start up something like Youtube or a local video file and play full screen, I notice my bandwidth consumption is 5-6mbps. When adding additional sessions that might be doing similar workloads, would I need to plan on similar metrics or would there be some additional compression and consolidating of traffic through the VPX10 or larger for sizing purposes? I am trying to guage what my next size option would be for bandwidth planning while not going too large.
@Rody Kossen I have made significant strides in getting the graphics polished and working on all platforms! I went back to the drawing board and started adding policies one by one and testing. Took some time but I think I have a winner! 🙂. One thing that I have noticed and found to be a common issues on several discussion boards is cursor black boxes/ghosting in Adobe suites when I enable “optimize for 3d workload” in my policies. This one policy alone is responsible for a much improved experience, but when I turn it on I get the black box. When I turn it off the box goes away. I have read about this dealing with 32bit cursors, etc. Is there a simple fix to apply that anyone knows about? I tried to add other adhoc policies to get the same performance from the citrix policies like full screen hardware codec, etc., but none get me to the same level of performance. I have noticed that if I logoff each time and get a new session, I don’t experience this issue. This seems to only occur when launching session on one platform, i.e. iPad, and then joining the session again from another platform like IGEL/Windows/MacBook. Not a major issue, just wondering if there is a known work around for such an experience?
3 Views
Open in Slack
PreviousNext
Powered by