Slurm HRES
5 steps1 pieces5 Sources
Slurm 26.11 introduces Hierarchical Resources (HRES) plugins to natively schedule quantum processing units (QPUs) as cluster-level resources, addressing structural flaws in previous models like generic resource binding and license-based idling. NVIDIA announced this expansion on November 16, 2026, enabling supercomputing centers to integrate QPUs into existing workflows without custom code. The HRES interface allows administrators to define QPU capacity and associated node groups, ensuring jobs wait in the queue if a device is unavailable rather than holding compute nodes idle. Community projects QRMI and QDMI have already released plugins leveraging this API, with adoption by IBM, Pasqal, IQM, and others. This development matters because it standardizes hybrid quantum-classical scheduling, reducing fragmentation for users and vendors. Key disagreements or uncertainties remain regarding the maturity of the beta feature and specific site policies for resource sharing. Readers should watch for the official release of Slurm 26.11 later this month and further documentation updates from SchedMD.
How it got here
On November 16, 2026, NVIDIA announced that Slurm 26.11 will include HRES plugins to support quantum computing integration, citing nearly sevenfold growth in QPU deployments at supercomputing centers over five years. [https://blogs.nvidia.com/blog/slurm-quantum-computing-supercomputing]
blogs.nvidia.com ↗On November 17, 2026, an NVIDIA technical blog detailed how HRES models QPUs as native resources, overcoming limitations of GRES and license-based approaches that caused node idling and upgrade risks. [https://developer.nvidia.com/blog/integrating-qpus-into-supercomputing-centers-using-slurm-hres-plugins]
developer.nvidia.com ↗