NCP-AII Question 144
Select 3You are tasked with deploying an AI workload on a cloud-native stack using NVIDIA GPU resources. To ensure efficient orchestration and scaling, you decide to use Kubernetes with the NVIDIA GPU Operator. Which of the following steps are required to correctly deploy and manage the GPUs in this environment?
- A
Install the NVIDIA GPU Operator in the Kubernetes cluster.
- B
Configure the NVIDIA Container Toolkit on each node manually before running workloads.
- C
Label the Kubernetes nodes with available GPU resources.
- D
Ensure the NVIDIA device plugin is running for GPU resource allocation.
- E
Manually schedule workloads to nodes with GPUs by specifying node names in the pod specification.
Show answer and explanation
Correct answers: A, C, D
Explanation
Deploying and managing GPUs in a cloud-native stack with Kubernetes and the NVIDIA GPU Operator requires specific steps. The GPU Operator automates the installation of key components like the NVIDIA device plugin and container toolkit, streamlining the process. Labeling nodes and ensuring the device plugin is running are critical for Kubernetes to correctly allocate GPU resources. Manual setup or scheduling is generally unnecessary in this configuration.
- A. Correct.
Correct. The NVIDIA GPU Operator is essential for managing GPU resources in a cloud-native Kubernetes environment. It automates the deployment of all necessary NVIDIA components.
- B. Incorrect.
Incorrect. The NVIDIA GPU Operator automates the configuration of the NVIDIA Container Toolkit, so manual setup is not required.
- C. Correct.
Correct. Labeling Kubernetes nodes with GPU resources is necessary to help the scheduler identify where GPU workloads can run.
- D. Correct.
Correct. The NVIDIA device plugin is required to allocate GPU resources for containers and ensure proper scheduling.
- E. Incorrect.
Incorrect. Kubernetes has a scheduler that automatically assigns workloads to nodes based on resource availability, so manual scheduling is not typically required.