One small question here … for a GPU based workload (data science, AI/ML, …) what would you recommend ?
Charmed Openstack or Charmed Kubernetes ?
What would be the easiest/safest to deploy and maintain for that purpose ?
Kind regards.
I would think it depends heavily on what applications you want to run. Kubeflow on top of Kubernetes or a different AI/ML package that is focused on running on VMs.
You can certainly also run Charmed Kubernetes on top of Charmed Openstack (you need to get the machines from somewhere), though then you start getting more difficulties accessing the raw hardware, as you have multiple layers of passthrough to navigate.
Thanks, but I was more thinking on thoughts about ease of deployment, day to day maintenance and performances.
I would think that deploying GPU based workload in Charmed Kubernetes on bare metal would provide the best experience but maybe I’m wrong.