Ubuntu 22.04 LTS for NVIDIA® GPU clusters (CUDA® 12)

One of the most popular Linux distributions and the most popular cloud OS. Used as the underlying system of many services, including Wikipedia, Uber, Instagram and many others. This is thanks to its scalability and compatibility with various software. Ubuntu is an open source OS. It is developed jointly by Canonical and the developer community. Long-term support for Ubuntu 22.04 LTS is valid until 2027.

The image is designed for Nebius AI virtual machines (VMs) with NVIDIA^® GPUs grouped into GPU clusters. It contains:

• NVIDIA Data Center Driver compatible with all GPUs available in Nebius AI.

• CUDA^® Toolkit 12, a development environment for creating high performance GPU-accelerated applications.

• MLNX_OFED drivers for InfiniBand adapters that are used to cluster VMs with GPUs.

1. Install and configure the Nebius AI CLI if you haven’t done it yet.

2. Create a GPU cluster:

   ncp compute gpu-cluster create \
     --name my-gpu-cluster \
     --interconnect-type infiniband
   
   

3. Create a VM with GPUs and add it to the GPU cluster using the --gpu-cluster-name flag:

   ncp compute instance create \
     --zone eu-north1-c \
     --platform gpu-h100 \
     --gpus 8 \
     --create-boot-disk image-folder-id=standard-images,image-family=? \
     --ssh-key=$HOME/.ssh/id_ed25519.pub \
     --network-interface subnet-id=<your_subnet_ID>,nat-ip-version=ipv4 \
     --gpu-cluster-name my-gpu-cluster
   
   

   The VM platform (--platform) must be gpu-h100 or gpu-h100-b (Intel Sapphire Rapids with NVIDIA^® Hopper^® H100), and the number of GPUs must be 8.

4. To add more VMs to the GPU cluster, create them using the CLI command above. All VMs in a cluster must use the same platform.

• High-performance computing for AI training and inference workloads, such as deep learning, neural networks, and machine learning.

• Parallel processing capabilities that accelerate scientific simulations, computational fluid dynamics, weather forecasting, and other computationally intensive tasks.

• Faster data exploration, pattern recognition, and predictive modeling valuable for industries that rely on big data analysis, such as finance, healthcare, and retail.

• Enhanced video transcoding, video analytics, and image recognition, enabling faster processing of multimedia content and supporting applications in fields such as media and entertainment.

Nebius AI does not provide technical support for the product. If you have any issues, please refer to the developer’s information resources.