Question

Your team runs a production Azure Kubernetes Service (AKS) cluster that must also host real-time inference endpoints managed from your Azure Machine Learning workspace. You need to make the existing AKS cluster available as a compute target in the workspace so that data scientists can deploy managed online endpoints against it, while ensuring the cluster remains usable for the team's other non-ML services. What should you do?

Accepted Answer

Install the Azure Machine Learning extension on the AKS cluster and attach it to the workspace as a Kubernetes compute target.

Answer

Create a new inference cluster of type AKS directly from the workspace, replacing the existing cluster.

Answer

Register the AKS cluster as a datastore in the workspace and point the online endpoint deployment to that datastore.

Answer

Create a compute instance in the workspace and configure it to forward inference traffic to the AKS cluster.