Artificial Intelligence Computing Leadership from NVIDIA
vCS - virtual compute server guide and/or help needed
Hey, maybe I'm blind and I'm for sure new to this at all, but is there a documentation how to setup a virtual compute server? Is it even possible with the trial access? I thought, signing up for the 90 days trial will give me access to all the software and drivers I need which will also provide a needed documentation to setup up everything. First problem (maybe just for my understanding) there is no explicit option for "compute server", just nutanix, redhat kvm, linux kvm, vsphere and citrix, which are virtualization products. So does the vCS only work with virtualization? Then my question would be, what's the difference between grid and vCS. I know, i can also have a graphic output for grid but I am not limited to it and can compute also. I thought, I maybe can use a vGPU for docker or cuda applications like tensorflow-gpu, etc.? I already installed the license server and uploaded a license file for Quadro-Virtual-DWS, which includes vCS, as far as i know. By the way, what is the difference between Linux KVM and RedHat KVM? In the official documents is the only supported KVM method RedHat KVM. But there are two different download products. I also just give it a "try and error" run and installed the linux-kvm stuff on centos and gentoo. (I don't have a RedHat installed and didn't plan to use it. (again)) Both of them give me this error "error: failed to send vGPU configuration info to RM: 6" What is RM? RedHat Manager? Does it only work with RedHat? Or is there a way to fix this? Maybe someone can shed light on my problems? Background: I'm working on a project to make our lecture (university) python/tensorflow, etc. online usable. At the moment we have some RTX 8000, RTX 6000 and some more are included in our ordering pipeline. So I already have some cards to test with and it's more or less a ground base for upcoming purchase decisions. Big thanks in advance and Greetings Matthias PS: On my wishlist would be a generic (linux) vgpu driver, which provides the host with vgpu devices which can be used by any application as long as the driver can get an active license from the license server. Don't know if it is applicable with the linux kernel. But I guess technically it is.
Hey,

maybe I'm blind and I'm for sure new to this at all, but is there a documentation how to setup a virtual compute server? Is it even possible with the trial access?

I thought, signing up for the 90 days trial will give me access to all the software and drivers I need which will also provide a needed documentation to setup up everything.
First problem (maybe just for my understanding) there is no explicit option for "compute server", just nutanix, redhat kvm, linux kvm, vsphere and citrix, which are virtualization products.
So does the vCS only work with virtualization? Then my question would be, what's the difference between grid and vCS. I know, i can also have a graphic output for grid but I am not limited to it and can compute also.
I thought, I maybe can use a vGPU for docker or cuda applications like tensorflow-gpu, etc.?
I already installed the license server and uploaded a license file for Quadro-Virtual-DWS, which includes vCS, as far as i know.

By the way, what is the difference between Linux KVM and RedHat KVM?
In the official documents is the only supported KVM method RedHat KVM.
But there are two different download products.

I also just give it a "try and error" run and installed the linux-kvm stuff on centos and gentoo.
(I don't have a RedHat installed and didn't plan to use it. (again))
Both of them give me this error "error: failed to send vGPU configuration info to RM: 6"
What is RM? RedHat Manager? Does it only work with RedHat?
Or is there a way to fix this?

Maybe someone can shed light on my problems?

Background:
I'm working on a project to make our lecture (university) python/tensorflow, etc. online usable.
At the moment we have some RTX 8000, RTX 6000 and some more are included in our ordering pipeline.
So I already have some cards to test with and it's more or less a ground base for upcoming purchase decisions.

Big thanks in advance and Greetings
Matthias

PS:
On my wishlist would be a generic (linux) vgpu driver, which provides the host with vgpu devices which can be used by any application as long as the driver can get an active license from the license server. Don't know if it is applicable with the linux kernel. But I guess technically it is.

#1
Posted 09/15/2020 12:25 PM   
Scroll To Top

Add Reply