virtual gpu development via cuda on linux help cuda on virtual cluster

Dear Sirs,
I intend on getting a cluster in the distant future but at the moment I am not ready for one thus i want to learn how to develop under a virtual cluster via User Mode Linux. Now I am at the moment setting up the cluster and learning about virtual machines and virtual networking and havent got around to installing CUDA yet but my question is this I have a Dell XPS with 2GB ram and a nvidia GPU , what i was to do is setup a virtual environment with 3 nodes cluster host node1 node 2 now I want to learn about parallel program execution and GPU parallel execution so is it possible on my virtual machines to setup virtual GPU’s one for each node and install cuda on each node and program as if i had a 3 node cluster just for learning purposes.

I understand that this setup will be incredibly slow but I have to start somewhere and funding does not allow me to run out and buy a Beowulf so can anyone point me in the right direction in order for me to reach my goal of a virtual networked GPU system on a virtual clustered machine development environment?

if you can help in any way then i will be delighted to hear from you.

Regards
Zoro

Hey,

I had some success in emulating CUDA code inside a VM in VMware Fusion. I’ve attached a copy of the installation instructions to this post. Good luck with your endeavor .
CUDA_Linux_Setup.mm.html (11.2 KB)

Thank you for your reply.

Regards

Zoro