|

| How to Integrate NVIDIA GPU Cloud with Microsoft Azure

How to Integrate NVIDIA GPU Cloud with Microsoft Azure

January 24, 2025

Discover how to seamlessly integrate NVIDIA GPU Cloud with Microsoft Azure for enhanced computing power and efficient AI workflows in this comprehensive guide.

How to Connect NVIDIA GPU Cloud to Microsoft Azure: a Simple Guide

Set Up and Configure Your Azure Account

Ensure you have an active Microsoft Azure account. You can sign up at the Azure website if you don't have one.

Access the Azure portal by navigating to https://portal.azure.com/.

In the Azure portal, ensure you have sufficient permissions to create resources and manage virtual networks.

Install Azure CLI

Download and install the Azure CLI from here.

After installation, open a terminal or command prompt and login using your Azure credentials:

az login

Register NVIDIA GPU Cloud (NGC)

Create an NVIDIA NGC account at NGC Signup.

After registration, verify your email and login to the NGC website.

Navigate to the 'API Key' section under your account settings. Generate and save this API key for accessing NGC resources.

Create an Azure Resource Group

In the Azure portal or using the CLI, create a new resource group to organize your NVIDIA GPU Cloud resources:

az group create --name myResourceGroup --location eastus

Deploy NVIDIA GPU-Optimized VM

Use the Azure Marketplace to search for NVIDIA GPU-optimized VMs or use the CLI to deploy a Deep Learning VM.

Run the following command to create a VM instance with an NVIDIA GPU:

az vm create \
  --resource-group myResourceGroup \
  --name myNvidiaVM \
  --image microsoft-dsvm:ubuntu-hvm:ubuntu-1804:latest \
  --size Standard_NC6

Install NVIDIA Drivers and Docker on Azure VM

SSH into the VM once it's running:

ssh azureuser@your-vm-ip-address

Update the package list and install the NVIDIA driver:

sudo apt-get update
cuda-drivers

Install Docker:

sudo apt-get install docker.io
sudo systemctl start docker
sudo systemctl enable docker

Configure Docker for NVIDIA GPU Support

Install the NVIDIA Docker container runtime:

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update

sudo apt-get install -y nvidia-docker2
sudo systemctl restart docker

Verify the installation using the NVIDIA Docker tool:

sudo docker run --rm --gpus all nvidia/cuda:11.0-base nvidia-smi

Access NVIDIA GPU Cloud CLI

Install the NGC CLI tool:

wget -O ngccli_cat_linux.zip https://ngc.nvidia.com/downloads/ngccli_cat_linux.zip && unzip ngccli_cat_linux.zip && rm ngccli_cat_linux.zip

Configure the NGC CLI with your API key obtained from the NGC website:

ngc config set

Enter your API key and set the default storage to Azure.

Pull and Run NVIDIA Docker Container

Use the NGC CLI to pull a Docker container with the desired framework:

ngc registry image pull nvcr.io/nvidia/tensorflow:21.03-tf1-py3

Run the pulled Docker container on your VM:

sudo docker run --gpus all -it nvcr.io/nvidia/tensorflow:21.03-tf1-py3

Omi Necklace

The #1 Open Source AI necklace: Experiment with how you capture and manage conversations.

Build and test with your own Omi.

How to Use NVIDIA GPU Cloud with Microsoft Azure: Usecases

Use NVIDIA GPU Cloud and Microsoft Azure for Deep Learning

Combine Forces: Leverage the power of NVIDIA GPU Cloud (NGC) with Microsoft Azure to build scalable deep learning models. This integration offers developers access to NGC's pre-trained models and AI software with Azure's vast computing resources and simplified machine learning tools.

Optimize Workflows: Utilize NVIDIA's GPU-optimized containers within Azure to streamline the process of training AI models. The containers are fully managed, offering a plug-and-play approach for data scientists aiming to leverage powerful GPU resources without the hassle of system maintenance.

Extreme Scalability: Azure's flexible architecture complements NGC's high performance. Deploying machine learning models on Azure using NVIDIA's technology ensures low latency and quick scaling abilities to handle varying workloads effectively.

Cost Efficiency: Both platforms allow usage-based billing, which can significantly reduce the cost of experimentation. Using Azure's spot instances with NVIDIA's scalable GPU solutions can optimize expense management.

Implementation Steps

Create an Azure Account: Sign up for Microsoft Azure and set up your subscription plan to begin leveraging Azure's cloud capabilities.

Access NVIDIA GPU Cloud: Register to access NVIDIA NGC and explore the curated set of software libraries, models, and tools specifically designed for AI, HPC, and data analytics.

Deploy GPU Instances in Azure: Utilize Azure's portal to create new virtual machines integrating NVIDIA GPUs. You can select from various options depending on the computational needs and budget.

Integrate NGC Containers: Pull relevant NGC containers into your Azure environment. These containers are equipped with the necessary software and libraries, easing setup procedures.

Conduct Experiments: Start training your deep learning models using TensorFlow, PyTorch, or any supported framework in the easy-to-deploy environment provided by NGC on Azure's infrastructure.

Monitor and Optimize Performance: Constantly assess your models' performance using Azure's monitoring tools and make necessary adjustments for better efficiency.

Code Example


docker pull nvcr.io/nvidia/tensorflow:<tag>

You can deploy this to your Azure-powered container on a GPU-enabled VM to start using NVIDIA's optimized TensorFlow framework.


az vm create --resource-group yourResourceGroup --name yourVM --image UbuntuLTS --size Standard_NC6 --admin-username azureuser

This command is an example illustrative of how you can create a VM in Azure optimized with NVIDIA GPUs, thereby boosting deep learning workflows when used in conjunction with NVIDIA NGC.

Transforming Healthcare AI with NVIDIA GPU Cloud and Microsoft Azure

Enhanced Diagnostics: Utilize NVIDIA GPU Cloud's (NGC) AI models in combination with Microsoft Azure's robust infrastructure to improve diagnostic accuracy. Leverage pre-trained imaging models from NGC on Azure’s AI-driven platform to facilitate faster and more accurate patient diagnostics.

Personalized Treatment: Apply deep learning models from NGC on Azure to tailor personalized treatment plans. By processing vast amounts of healthcare data on Azure GPU instances, AI can facilitate insights into specific patient needs, improving treatment outcomes.

Data Security and Compliance: With Azure's compliance certifications and NVIDIA’s secure containerized environments, manage patient data securely across AI workflows, adhering to critical health regulations such as HIPAA.

Healthcare AI Scalability: Azure's flexible scaling options allow healthcare providers to adjust resources as needed, testing various models from NGC. This helps tackle computational challenges during peak data processing hours effectively.

Implementation Steps

Initiate Azure Setup: Create a Microsoft Azure account to unlock cloud resources essential for handling healthcare applications efficiently.

Explore NVIDIA GPU Cloud Offerings: Register on NGC to access enhanced AI tools and medical imaging models optimized for Microsoft Azure environments.

Provision GPU Instances: Utilize Azure to set up GPU-powered virtual machines, specifically optimizing for workloads requiring substantial computation, like imaging diagnostics.

Deploy and Run NGC Containers: Use Azure’s container services to run NVIDIA's AI containers, thus integrating potent AI models into healthcare applications seamlessly.

Evaluate AI Models: Train and test advanced AI models for predictive analytics using Azure's machine learning operations (MLOps) tools to refine patient diagnosis systems.

Maintain Regulatory Standards: Employ Microsoft Azure's compliance tools to continually audit AI applications, ensuring they meet necessary healthcare regulation benchmarks.

Code Example

docker pull nvcr.io/nvidia/clara-train-sdk:<version>

This command pulls the NVIDIA Clara Train SDK, designed to enhance medical imaging tasks, which can be deployed directly onto a GPU-enabled Azure VM for advanced AI capabilities.

az vm create --resource-group healthcareGroup --name healthcareVM --image UbuntuLTS --size Standard_NC12 --admin-username azureuser

This Azure CLI command illustrates setting up a virtual machine that supports NVIDIA GPU, crucial for running complex healthcare AI models hosted on NGC, thus aiding in transformative patient care solutions.