Skip to content

Overview

Rafay now offers seamless integration with NVIDIA BlueField-3 Data Processing Units (DPUs), enabling turnkey orchestration of high-performance, secure, and multi-tenant GPU Kubernetes environments. Traditionally complex to deploy, BlueField-enabled infrastructure now becomes fully automated and repeatable using Rafay's custom DPF Blueprint.

This integration abstracts away the multi-step setup of NVIDIA’s DOCA and DPF stack—including the CNI, DPU control plane, and security services—into a single-click deployment using Rafay’s PaaS platform. Platform teams can easily scale and manage AI-ready clusters with built-in security, lifecycle automation, and observability—without requiring deep kernel or networking expertise.

Rafay’s BlueField Blueprint supports:

  • Automatic setup of NVIDIA DOCA/DPF components

  • OVN-Kubernetes CNI acceleration

  • Offloading of networking, storage, and security functions to the DPU

  • Multi-tenant routing, encryption, and GPU workload optimization

This makes it the fastest, most reliable way to operationalize DPU-accelerated AI infrastructure in Kubernetes at scale.

DPF Blueprint

Step 1: Download RCTL

The RCTL CLI allows you to programmatically interact with the controller enabling users to construct sophisticated automation workflows.

  • Login into your Org
  • Navigate to My Tools to download both the RCTL CLI and the CLI Config file
  • Initialize RCTL using the step-by-step instructions
  • Ensure you update your OS's Path environment variable for RCTL

Step 2: Download Files

You will download a ZIP archive containing specification files that will be used by RCTL to create the needed resources within the controller.

Step 3: Create Namespaces

In this step, you will use RCTL to create all of the namespaces needed for the cluster blueprint add-ons.

Info

The specifications files are configured for a project named dpf-test. If you are using a different project name, the specification files will need to be updated.

  • In your CLI, navigate to the /dpu/namespaces in your extracted archive
  • Run the following command to loop through all of the YAML files and create the respective namespaces
for file in ./*.yaml; do [ -f "$file" ] && rctl apply -f "$file"; done
  • In your project, navigate to Infrastructure -> Namespaces where you will see the newly created namespaces

Create Namespaces

Step 4: Create Repositories

In this step, you will use RCTL to create all of the repositories needed for the cluster blueprint add-ons.

Info

The specifications files are configured for a project named dpf-test. If you are using a different project name, the specification files will need to be updated.

  • In your CLI, navigate to the /dpu/repositories in your extracted archive
  • Run the following command to loop through all of the YAML files and create the respective repositories
for file in ./*.yaml; do [ -f "$file" ] && rctl apply -f "$file"; done
  • In your project, navigate to Integrations -> Repositories where you will see the newly created repositories

Create Repos

Step 5: Create Add-Ons

In this step, you will use RCTL to create all of the Add-On needed for the cluster blueprint.

Info

The specifications files are configured for a project named dpf-test. If you are using a different project name, the specification files will need to be updated.

  • In your CLI, navigate to the /dpu/addons in your extracted archive
  • Run the following command to loop through all of the YAML files and create the respective add-ons
for file in ./*.yaml; do [ -f "$file" ] && rctl apply -f "$file"; done
  • In your project, navigate to Infrastructure -> Add-Ons where you will see the newly created add-ons

Create addons

Step 6: Create Cluster Blueprint

In this step, you will use RCTL to create the cluster blueprint.

Info

The specifications files are configured for a project named dpf-test. If you are using a different project name, the specification files will need to be updated.

  • In your CLI, navigate to the /dpu/blueprint in your extracted archive
  • Run the following command to create the blueprint
rctl apply -f dpf-rafay-blueprint.yaml
  • In your project, navigate to Infrastructure -> Blueprints where you will see the newly created blueprint

Create blueprint

Recap

You have successfully created a custom cluster blueprint containing the add-ons needed for NVIDIA BlueField-3 DPU Integration. This blueprint can be applied to a fleet of Kubernetes clusters.