split GPU for multiple users

[#1]

Hi everyone:

CDSW support using GPU for machine learning jobs, but it seems one GPU can only be used by one user at the same time.

Can we share one GPU with mulitple users or one GPU must be dedicated to one docker and nobody else can use it at the same time?

Thanks！

[#2 · Accepted solution]

Hi,

CDSW relies on Kubernetes to allocate system resources to user workloads and Kubernetes doesn't allow sharing a GPU core like it does for CPUs.

The expectations are set in the Kubernetes documentation:

• Containers (and pods) do not share GPUs. There’s no overcommitting of GPUs.
• Each container can request one or more GPUs. It is not possible to request a fraction of a GPU.

from: https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/

Regards,

Peter

[#3]

Hi, Peter:
Thank you for your reply. That's really helpful.