Prerequisites

NVIDIA AI Enterprise 2.0 or later

Hardware Requirements

NVIDIA AI Enterprise with Red Hat OpenShift and VMware vSphere has the following sets of prerequisites:

• At least three NVIDIA AI Enterprise Compatible servers that are NVIDIA-Certified System.

• At least one of NVIDIA AI Enterprise Compatible servers must have a NVIDIA AI Enterprise supported NVIDIA GPU.

• Recommended A100 for training and A30 for inference.

  • Single Root I/O Virtualization (SR-IOV) Enabled

  • VT-d/IOMMU Enabled

• The GPU accelerated server(s) must have NVIDIA AI Enterprise Host Software (VIB) installed.

  Note

  The installation of VMware ESXi and the NVIDIA vGPU Host and Guest Driver Software is out of the scope of this document. Please refer to the NVIDIA AI Enterprise Virtualization Deployment Guide for detailed instructions. To set up AI-ready VMs on VMware, a vGPU profile needs to be added to the VM. This requires installing the vGPU Host Manager on ESXi, attaching a vGPU profile, installing a vGPU guest driver on the VM, and licensing the VM. The following sections of the guide are helpful for reference:

  • Installing VMware ESXi

  • Installing and Configuring NVIDIA AI Enterprise Host Software

  • Deploying the NVIDIA License System

  • Creating a VM and installing the NVIDIA Driver in the VM

  • Selecting the Correct vGPU Profile

Once the three NVIDIA AI Enterprise Compatible servers have met the above NVIDIA AI Enterprise hardware and software requirements, you must choose a method to install OpenShift Container on vSphere. For the authoring of this document, the Installer-provisioned infrastructure (IPI), was chosen since it is pre-configured and automates the provisioning of resources which are required by OpenShift Container Platform.