Elastic Resource Quota for non-AI workloads

[kaiohenricunha]

I operate a regular DevOps platform for microservices which doesn't use or need GPU.

However, GPU is a prerequisite.

The Elastic Resource Quota feature really caught my attention. It would be a great help.

Is there a way to use it without having to enable GPU?

[Telemaco019]

Hi @kaiohenricunha, thanks for your interest in nos! The Elastic Resource Quota feature is an extension of the Capacity Scheduling plugin from the kubernetes-sig/scheduling.

If you want to use Elastic Resource Quotas without GPU support, you can either use the Capacity Scheduling plugin mentioned above or install nos disabling the GPU Partitioner component:

helm install oci://ghcr.io/nebuly-ai/helm-charts/nos \
  --version 0.1.2 \
  --namespace nebuly-nos \
  --generate-name \
  --create-namespace \ 
  --set gpuPartitioner.enabled=False 

However, as nos is meant for managing GPU workloads, if your cluster doesn't use GPUs then I'd recommend going with the original Capacity Scheduling plugin. I hope this helps!

[kaiohenricunha]

Hi @kaiohenricunha, thanks for your interest in nos! The Elastic Resource Quota feature is an extension of the Capacity Scheduling plugin from the kubernetes-sig/scheduling.

If you want to use Elastic Resource Quotas without GPU support, you can either use the Capacity Scheduling plugin mentioned above or install nos disabling the GPU Partitioner component:

helm install oci://ghcr.io/nebuly-ai/helm-charts/nos \
  --version 0.1.2 \
  --namespace nebuly-nos \
  --generate-name \
  --create-namespace \ 
  --set gpuPartitioner.enabled=False 

However, as nos is meant for managing GPU workloads, if your cluster doesn't use GPUs then I'd recommend going with the original Capacity Scheduling plugin. I hope this helps!

I think I still prefer using Nos because of its ease of installation(Helm)