Artificial Intelligence Computing Leadership from NVIDIA
Tesla M10 ESX Windows RDSH Question
Hello, So I purchased 3 new Fujitsu servers equipped with the Tesla M10 and have ESX V6.7 running on all of them. After the initial purchase I realized that licenses where needed. My plan is to install Windows 2019 Remote Desktop Services strictly for use with Wyse Thin Clients doing RDSH only. There will be one RDS server installed on each ESX server, my hope was to just give our remote sessions a little boost in graphic performance cause in our current 2008 environment we get a bit of screen scraping. From reading in the forums it appears my best setup will be doing Shared Direct and adding the PCI device to the guest (Pass Through). Across all 3 servers there will be a total of around 150 workstations that connect via RDSH. What kind of licensing will this require and is my approach proper? Thanks!
Hello,

So I purchased 3 new Fujitsu servers equipped with the Tesla M10 and have ESX V6.7 running on all of them.
After the initial purchase I realized that licenses where needed.
My plan is to install Windows 2019 Remote Desktop Services strictly for use with Wyse Thin Clients doing RDSH only.
There will be one RDS server installed on each ESX server, my hope was to just give our remote sessions a little boost in graphic performance cause in our current 2008 environment we get a bit of screen scraping.
From reading in the forums it appears my best setup will be doing Shared Direct and adding the PCI device to the guest (Pass Through).
Across all 3 servers there will be a total of around 150 workstations that connect via RDSH.
What kind of licensing will this require and is my approach proper?

Thanks!

#1
Posted 10/15/2019 04:42 PM   
Hi, not sure how this should work properly with 1 RDSH VM per physical host. You would need at least 4 RDSH VMs per host to leverage the Tesla M10 (4GPUs in Passthrough each). For your scenario you will need 150 vApps licenses if these 150 Users connect concurrently. Otherwise you will need the number of CCU. Hope this helps. Regards Simon
Hi,
not sure how this should work properly with 1 RDSH VM per physical host. You would need at least 4 RDSH VMs per host to leverage the Tesla M10 (4GPUs in Passthrough each).
For your scenario you will need 150 vApps licenses if these 150 Users connect concurrently. Otherwise you will need the number of CCU.

Hope this helps.

Regards
Simon

#2
Posted 10/15/2019 07:33 PM   
Simon, Correct, I would just leverage a single GPU per VM Guest RDSH which my hope is that would be much better than using the standard built in VMWare Graphics adapter and take some of the load off other resources. Is this reasoning correct? The reason for putting a single RDSH on each server is for redundancy (I won't go into the whole enviroment and how it is setup). There is currently about 35 other servers spread across the deployment. We probably over purchased with the Tesla M10 considering our environment but I basically asked our provider to put a video card that could be leveraged for a RDS Guest in esx that was fully supported in this configuration and that is what was provided. Since it is on 3 different esx host and we have up to 150 user connect concurrently would the vApps cover all 3 NVIDIA M10 cards on the machines?
Simon,

Correct, I would just leverage a single GPU per VM Guest RDSH which my hope is that would be much better than using the standard built in VMWare Graphics adapter and take some of the load off other resources.
Is this reasoning correct?
The reason for putting a single RDSH on each server is for redundancy (I won't go into the whole enviroment and how it is setup). There is currently about 35 other servers spread across the deployment.

We probably over purchased with the Tesla M10 considering our environment but I basically asked our provider to put a video card that could be leveraged for a RDS Guest in esx that was fully supported in this configuration and that is what was provided.

Since it is on 3 different esx host and we have up to 150 user connect concurrently would the vApps cover all 3 NVIDIA M10 cards on the machines?

#3
Posted 10/15/2019 08:35 PM   
Hi Just so we understand this correctly, as you have "[i]about 35 other servers spread across the deployment[/i]", your intention is to give your users a better graphical experience (on 3 RDSH VMs) and you'll use these 3 RDSH VMs to add resilience in your entire deployment. You intend to try and support all 150 concurrent users using (only) 3 RDSH VMs? ... Apologies in advance, but that does need some clarity, because (as Simon mentions above) 3 RDSH VMs with an 8GB GPU will not support 150 concurrent users. [quote=""]I would just leverage a single GPU per VM Guest RDSH which my hope is that would be much better than using the standard built in VMWare Graphics adapter and take some of the load off other resources. Is this reasoning correct?[/quote]Yes [quote=""]We probably over purchased with the Tesla M10 considering our environment but I basically asked our provider to put a video card that could be leveraged for a RDS Guest in esx that was fully supported in this configuration and that is what was provided.[/quote] We can look in to this when you've confirmed the first question at the top :-) [quote=""]Since it is on 3 different esx host and we have up to 150 user connect concurrently would the vApps cover all 3 NVIDIA M10 cards on the machines?[/quote] The NVIDIA License Server sits centrally to the deployment. All RDSH VMs will need to be able to contact the NVIDIA License Server. By doing so, all user sessions will be licensed correctly. Again, you mention 150 concurrent users, 3 RDSH VMs will not be able to support this. Regards MG
Hi

Just so we understand this correctly, as you have "about 35 other servers spread across the deployment", your intention is to give your users a better graphical experience (on 3 RDSH VMs) and you'll use these 3 RDSH VMs to add resilience in your entire deployment. You intend to try and support all 150 concurrent users using (only) 3 RDSH VMs? ...

Apologies in advance, but that does need some clarity, because (as Simon mentions above) 3 RDSH VMs with an 8GB GPU will not support 150 concurrent users.

said:I would just leverage a single GPU per VM Guest RDSH which my hope is that would be much better than using the standard built in VMWare Graphics adapter and take some of the load off other resources.
Is this reasoning correct?
Yes

said:We probably over purchased with the Tesla M10 considering our environment but I basically asked our provider to put a video card that could be leveraged for a RDS Guest in esx that was fully supported in this configuration and that is what was provided.
We can look in to this when you've confirmed the first question at the top :-)

said:Since it is on 3 different esx host and we have up to 150 user connect concurrently would the vApps cover all 3 NVIDIA M10 cards on the machines?
The NVIDIA License Server sits centrally to the deployment. All RDSH VMs will need to be able to contact the NVIDIA License Server. By doing so, all user sessions will be licensed correctly. Again, you mention 150 concurrent users, 3 RDSH VMs will not be able to support this.

Regards

MG

#4
Posted 10/21/2019 08:24 AM   
Scroll To Top

Add Reply