Red Hat Expands AI Infrastructure Capabilities with Support for NVIDIA BlueField DPUs

Red Hat, provider of open source solutions, has announced support for Red Hat OpenShift on NVIDIA BlueField data processing units (DPUs), creating a more secure and high-performance platform for AI-driven workloads and next-generation data centers.

Modern enterprises face growing challenges in balancing the compute demands of AI applications with the networking and security services that support them. Red Hat’s latest integration addresses these challenges by offloading infrastructure tasks from CPUs to DPUs, freeing up resources for AI workloads while improving overall system efficiency and security.

The joint solution enables:

  • Optimized resource utilization by shifting networking tasks from the CPU to the DPU.
  • Accelerated data and storage performance through support for NVMe over Fabrics (NVMe-oF) and accelerated OVS data paths.
  • Enhanced cloud networking with distributed routing for better scalability and traffic efficiency.
  • Improved security posture by isolating application workloads from infrastructure services.

Support for NVIDIA BlueField on Red Hat OpenShift will roll out as a technical preview in the coming weeks, with deeper integration planned for NVIDIA’s DOCA software framework and third-party network functions.

Looking ahead, NVIDIA BlueField-4 is expected to extend these capabilities with next-generation acceleration, tighter DOCA integration, and enhanced performance to power large-scale AI factories. Red Hat is also collaborating with NVIDIA to support Spectrum-X Ethernet networking, enabling seamless, high-speed AI workloads across distributed cloud environments.

“Our collaboration with NVIDIA provides customers with a more reliable, secure, and high-performance platform to address the growing demands of AI workloads,” said Ryan King, Vice President, AI and Infrastructure, Partner Ecosystem Success, Red Hat.

“The integration of NVIDIA BlueField with Red Hat OpenShift marks a new era of secure and efficient infrastructure for data-intensive AI workloads,” added Justin Boitano, Vice President, Enterprise Products, NVIDIA.

Leave a Reply

Your email address will not be published. Required fields are marked *