NVIDIA Acquires SchedMD, Developer of the Slurm Workload Manager

NVIDIA Acquires SchedMD, Developer of the Slurm Workload Manager

NVIDIA has acquired SchedMD, the developer of Slurm, an open-source workload management system used in high-performance computing (HPC) and artificial intelligence. The acquisition is intended to support the open-source software ecosystem and advance AI development for researchers, developers, and enterprises.

NVIDIA said Slurm will continue to be developed and distributed as open-source, vendor-neutral software. It will remain available to the broader HPC and AI community and supported across a range of hardware and software environments.

Managing HPC and AI workloads

HPC and AI workloads perform complex computations by running parallel tasks across clusters. These environments require systems that can queue jobs, schedule tasks, and allocate computing resources efficiently. As clusters grow larger and more powerful, effective resource utilization becomes increasingly important.

Slurm is a workload manager and job scheduler designed for scalability, high throughput, and complex policy management. It is used by more than half of the top 10 and top 100 systems listed in the TOP500 ranking of supercomputers.

The software also supports the latest NVIDIA hardware and is used as part of the infrastructure for generative AI. Foundation model developers and AI builders use Slurm to manage model training and inference workloads.

Continued open-source development

SchedMD CEO Danny Auble said the acquisition recognizes Slurm's role in demanding HPC and AI environments. He added that NVIDIA's expertise and investment in accelerated computing would support Slurm's development for future AI and supercomputing requirements, while the software would remain open source.

NVIDIA and SchedMD have collaborated for more than a decade. NVIDIA said it will continue investing in Slurm and maintaining it as an open-source scheduler for HPC and AI.

The acquisition will give SchedMD access to new systems and is expected to help users of NVIDIA's accelerated computing platform optimize workloads across their compute infrastructure. NVIDIA also said it would continue supporting diverse hardware and software environments, enabling customers to operate heterogeneous clusters while using newer Slurm capabilities.

NVIDIA will continue providing open-source software support, training, and development services for SchedMD's hundreds of customers. These customers include cloud providers, manufacturers, AI companies, and research laboratories working in autonomous driving, healthcare and life sciences, energy, financial services, manufacturing, and government.

Through the acquisition, NVIDIA and SchedMD aim to support the continued use of open-source workload management across HPC and AI environments of different sizes and types.