NVIDIA
vGPU vs. Passthrough combined with VMware vSphere Standard vs. Enterprise Plus (M10)
For a Xenapp deployment based on 7.9 and M10 GPU's, would there be a valid reason for using vGPU instead of 4x passthrough? We are hosting 4 Citrix VM's per host. It would require expensive VMware licenses (enterprise plus instead of standard) to be able to use vGPU and I was wondering if there's anything that would warrant such a price difference.
For a Xenapp deployment based on 7.9 and M10 GPU's, would there be a valid reason for using vGPU instead of 4x passthrough? We are hosting 4 Citrix VM's per host.

It would require expensive VMware licenses (enterprise plus instead of standard) to be able to use vGPU and I was wondering if there's anything that would warrant such a price difference.

Hans Kraaijeveld, Senior Project Engineer, EUC Consultant, Cliënt ICT Groep
Twitter: @hanskraaijeveld
Blog: https://www.plonius.com

#1
Posted 08/28/2017 09:33 AM   
Hi, currently you would benefit from features like monitoring when using vGPU instead of Passthrough. In the future there will be additional features like motion support. So if you don't need the provided features you're fine with Passthrough and ESX Standard. Regards Simon
Hi,

currently you would benefit from features like monitoring when using vGPU instead of Passthrough. In the future there will be additional features like motion support. So if you don't need the provided features you're fine with Passthrough and ESX Standard.

Regards

Simon

#2
Posted 08/29/2017 10:08 AM   
Hi Simon, Thanks for the validation of my presumption. In a XenApp environment, there wouldn't be much use for vGPU motion support (very cool to watch happening though @VMworld!!! The running Heaven Benchmark didn't even flinch! #epic) grtz, Hans
Hi Simon,

Thanks for the validation of my presumption. In a XenApp environment, there wouldn't be much use for vGPU motion support (very cool to watch happening though @VMworld!!! The running Heaven Benchmark didn't even flinch! #epic)

grtz,

Hans

Hans Kraaijeveld, Senior Project Engineer, EUC Consultant, Cliënt ICT Groep
Twitter: @hanskraaijeveld
Blog: https://www.plonius.com

#3
Posted 08/30/2017 11:56 AM   
Hi Hans If you go down the Passthrough route, you will loose all new functionality when it's released through the GRID software updates. By not using vGPU (GRID), you effectively go down a hardware only route, where there is no feature development as it's all built into the GRID software. Sure, you have a Windows GPU driver, but that's it. Regarding not using vMotion for XenApp, how do you apply updates to your ESX hosts? Would it not be beneficial to be able to migrate your VMs to another ESX host without having to drain and power down your XenApp VMs to restart the host? I wouldn't base your decision on the current GRID 4.3 release, make sure you evaluate the GRID 5 release coming on September 1st. There's a lot of new features in this release that may make it worth while :-) Regards Ben
Hi Hans

If you go down the Passthrough route, you will loose all new functionality when it's released through the GRID software updates. By not using vGPU (GRID), you effectively go down a hardware only route, where there is no feature development as it's all built into the GRID software. Sure, you have a Windows GPU driver, but that's it.

Regarding not using vMotion for XenApp, how do you apply updates to your ESX hosts? Would it not be beneficial to be able to migrate your VMs to another ESX host without having to drain and power down your XenApp VMs to restart the host?

I wouldn't base your decision on the current GRID 4.3 release, make sure you evaluate the GRID 5 release coming on September 1st. There's a lot of new features in this release that may make it worth while :-)

Regards

Ben

#4
Posted 08/30/2017 01:15 PM   
Hi Ben, We make sure we have enough citrix servers rolled out to be able to handle the failure of an entire host in an N+1 scenario. For maintenance, we would do the same. Put the Citrix VM's in maintenance (remove from the delivery group), power off the Citrix VM's, perform maintenance, reboot the host and there we go. What would be the features that would make it worthwhile in your opinion? Grtz, Hans
Hi Ben,

We make sure we have enough citrix servers rolled out to be able to handle the failure of an entire host in an N+1 scenario. For maintenance, we would do the same. Put the Citrix VM's in maintenance (remove from the delivery group), power off the Citrix VM's, perform maintenance, reboot the host and there we go.

What would be the features that would make it worthwhile in your opinion?

Grtz,

Hans

Hans Kraaijeveld, Senior Project Engineer, EUC Consultant, Cliënt ICT Groep
Twitter: @hanskraaijeveld
Blog: https://www.plonius.com

#5
Posted 09/07/2017 08:36 AM   
Scroll To Top

Add Reply