Looking to try and eke some more user density out ...
# citrix-vad
j
Looking to try and eke some more user density out of an ageing Citrix environment running on Nutanix. Are there any decent links out there pointing to good Citrix policy configurations to try and get as many on as possible?
l
Have you got WEM? May give you a bit of mileage, if not, optimising the crap out of your images, making sure GPU acceleration is disabled across the board (unless you have them of course), making sure edge/chrome is tidying up tabs, can't remember what that's called now. MCS I/O / PVS RAM Cache could give you a bit of a boost too.
j
All good tips, cheers. I can remember somewhere there being a kind of guide document on Citrix policies to set to get the best possible performance (it was quite old I think), does anyone know if this is still around anywhere? My Google-wrangling or ChatGPT-prompting doesn't seem to be getting me there, although Citrix could well have binned the doc or locked it up behind a paywall
c
WEM or Ivanti performance manager optimization will help (CPU Spike protection/Reclaiming idle memory. Running Citrix optmizer, High performance power management settings on hosts, There is policy templates in studio/web studio for server scalability as well
j
Scalability, that was the search term I needed 🙂
Doh! Forgot about templates
c
image.png
👍 1
👍🏻 1
l
I must have a poke around in there and see what it's doing.
c
yeah got to be careful with it and review you can create a policy from the template and remove anything you dont want ....
it only autocreates the clients default printer and turns off multimedia conferencing so that may break alot of stuff in your enviroment
l
Might be interesting to test and see what sort of impact that has on scalability generally.
c
yeah visual quality lower , no video codec , lower fps might lower the cpu usage and bandwith a little but cant see it having that much of an effect tbh
l
I'll add it to a GO-EUC board to do some research.
👀 1
👍 1
c
Uberagent is included in UHMC/Platform license so would stick that on as well. You need a splunk backend or can use something like Elasticsearch & Kibana but dashboards are way better in splunk. That will allow you dive deep into performance metrics /User experience scores on the boxes
l
John raises a good point about agents. Make sure those exclusions are up to date and it's not persistent you are not scanning on reads on the VM level.
m
Honestly the biggest bang for buck, in no order, make sure your gold master is optimized (ensure high performance mode, GPU is working and Citrix policies are working for GPU acceleration), including snapchains of the GM if using MCS, ensure you're on a somewhat modern AHV (>6.5 as AHV8 had alot of optimizations), and ensure you're rightly sized so you're not committing the sin of overprovisioning or underprovisioning. Afterwards? Look and NUMA fragmentation. And don't forget about good old defrag, yes, it's still a thing, on Gold master.
Otherwise? I mean sure giving CVMs more memory CAN, help. But it depends. Like anything with RAM, more only helps if it needed it. Hurts you in the long run to just throw RAM if not needed.
Ensure your endpoint security is correctly bypassed for trusted services and files, that's a biggie. Don't care who your provider is, MDE, Trellix or CS
Otherwise? Eh, I've played with things like SleepStates, Dell BIOS optimizations, eh, they're kinda not worth it. Do, check your CPU temps to ensure your stuff isn't throttling if it's that old. If you're adverterous, and CPU heat is an issue? Repasting CPUs is a thing, but ensure that's actually your bottleneck (thermal constrained impacting Turbo Frequencies) before endeavoring down that route
Oh, although I personally advise against it for any kind of corporate / large org / federal engagement, which CVAD deployments typically are, for your SMB mom and pop? Don't forget on legacy Spectre and Meltdown mitigations have a compute "tax" you have to pay. It's worth testing disabling for low risk deployments where benchmarks show disabling mitigations help. I would only do this on host your own / in house deployments, I would not disable Spectre / Meltdown mitigations on anything "outhouse" deployed. I've found in my testing anything somewhat recent, IE last 10 years, doesn't help so much. But, if you're running something say Circa 2012? Yeah, I've found in certain cases, disabling can help if there's alot of context swapping
👍🏻 1
a
So disable graphics acceleration in Edge? Then put tabs to sleep to save CPU? if people are having lets say 100s of edge tabs up and running? We have it but I think 1 hour is too long, might lower it to 30 min..