AI-300 Practice Question 72: Kubernetes Compute on Prem

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 8 · question 2 of 10

Design and implement an MLOps infrastructuremedium

A company must train models on GPUs in its own on-premises Kubernetes cluster because of data residency rules, yet wants to submit and track jobs from its Azure Machine Learning workspace. What should the team do?

Select one answer
Show answer and explanation

Answer:

  • C. Attach the cluster as Kubernetes compute, using Azure Arc and the ML extension

Azure Machine Learning can use a Kubernetes cluster as a compute target. A cluster outside AKS is connected through Azure Arc, the Azure Machine Learning extension is installed, and the cluster is attached to the workspace. Jobs are then submitted from the workspace but run on the company's own GPUs (option C).

- A is wrong because the data would then be processed in Azure, which conflicts with the residency requirement. - B is wrong because serverless compute runs on Azure-managed machines, so training would still run outside the company's own cluster. - D is wrong because a compute instance is an Azure VM, so the data would leave the on-premises environment.

Reference: Introduction to Kubernetes compute target

Practice Test 8 ›

Study guides for AI-300