Deployment of a GPU-based container to a cluster

Hello, everyone.

I am new to Portainer and web-based application development. I am trying to get some high-level idea about if it is possible/how to use Portainer to deploy a containerized GPU-enabled simulation platform to a computing cluster (~5 Linux servers and each server is equipped with several NVIDIA GPUs) that:

  1. Allows each user to
  • Create a temporary user account;
  • Upload input files to the server and submit simulation jobs (created from a specific image with certain container settings);
  • Download simulation results from the server.
  1. Allows server developer/maintainer to
  • Control access of each user: for instance, users can only use provided images with certain settings;
  • Schedule jobs according to the availability of GPUs;
  • Kill jobs that exceeds a specific runtime threshold.

Any comments and suggestions will be appreciated.

Thank you!
Shijie

Hi
I am not sure if i undertsood the question correctly.Kindly follow this feature on GPU support with portainer,https://github.com/portainer/portainer/issues/3143. In slack there are some users who does work in GPu’s pl join there, may be we could better response.

Hi! Thank you so much for the feedback! I will post the GPU-related questions under that thread.