AMD and RGS Deliver Security-Focused HPC and AI on Kubernetes

News


AMD and Rancher Government Solutions unveil a hardened architecture to optimize HPC and AI workloads on Kubernetes, enhancing GPU utilization and security.



AMD and RGS Deliver Security-Focused HPC and AI on Kubernetes

AMD and Rancher Government Solutions (RGS) have announced a new partnership aimed at simplifying high-performance computing (HPC) and artificial intelligence (AI) workload scheduling on Kubernetes. The collaboration introduces a hardened, production-ready reference architecture designed to improve GPU utilization and meet stringent security requirements for sensitive workloads.

Optimizing HPC and AI with Kubernetes

The solution leverages the RGS Carbide Platform and AMD Instinct™ GPUs to provide a scalable foundation for GPU-accelerated workloads. By integrating open-source projects Slinky and Kueue, the architecture enables intelligent job scheduling and automated GPU allocation, ensuring peak efficiency for research and national security organizations.

Chris Reynolds, Vice President of Solution Architects at RGS, emphasized the importance of this development: “HPC and AI are converging quickly. As workloads grow, organizations need to share GPU resources efficiently while meeting high security standards. This reference architecture is built to do both. Because it runs on Kubernetes, researchers can develop on a single workstation and scale the same work to a large GPU cluster with a consistent baseline.”

Addressing Traditional HPC Challenges

Traditional HPC systems often face challenges such as slow job deployment and mismatched environments between researchers’ workstations and shared systems. The new architecture addresses these issues by utilizing containerization, which packages jobs with their code, libraries, and dependencies. Kubernetes then orchestrates these containers, allowing researchers to define job requirements declaratively, such as GPU count and memory allocation. This approach ensures reproducibility, reduces idle capacity, and improves hardware utilization.

The architecture also automates the deployment of the full stack, from GPU drivers to job schedulers, enabling organizations to set up identical environments repeatedly. This allows data scientists and researchers to focus on their work rather than infrastructure maintenance.

Hardened Security for Sensitive Workloads

Given the sensitive nature of many HPC environments, the architecture includes a secure, compliant baseline. Key features include:

  • Optimized Compute: High-performance AMD Instinct GPUs with automated installation of the AMD ROCm software stack.
  • Security-Focused Orchestration: RKE2 Government, a hardened Kubernetes distribution running on secure enterprise Linux, designed to meet U.S. Government security standards.
  • GPU Management: AMD ROCm GPU Operator for automatic GPU resource management in containerized workloads.
  • Hardened Baseline and Automation: The Carbide Platform provides a secure foundation with automated deployment from drivers to job scheduling.

Fair and Efficient GPU Scheduling

The architecture integrates Slinky and Kueue for effective GPU scheduling. Slinky brings the widely-used Slurm scheduler to Kubernetes, while Kueue manages job queues and quotas. This combination ensures fair resource sharing and prevents idle capacity, allowing researchers to run demanding applications such as high-throughput large language model inference.

Strategic Implications

This initiative builds on AMD’s leadership in accelerated computing, which includes powering exascale supercomputers and enabling large-scale AI inference. By combining AMD’s hardware and open ROCm software with RGS’s hardened Kubernetes foundation, the partnership aims to help government agencies, research institutions, and regulated enterprises deploy and operate GPU-accelerated infrastructure with greater confidence.

Share with your friends!

Products You May Like

Leave a Reply

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

Please enter CoinGecko Free Api Key to get this plugin works.