Slurm 26.11 Adds Plugin Support for Quantum Processor Scheduling
Supercomputing centers can account for shared QPU capacity and keep jobs queued until a device is available.

Slurm 26.11 introduces a plugin interface for scheduling quantum processing units (QPUs) as shared cluster resources, letting jobs wait in a queue until devices are available.
The interface uses Slurm’s Hierarchical Resources (HRES) system. HRES lets operators describe a resource’s capacity and the groups of compute nodes eligible to access it. The new plugin support connects that resource model to QPU scheduling within the job management system centers already use for conventional computing.
Operators can define resource details in configuration, while a Lua plugin can check device health. The plugin can handle job admission and resource release. If the device is unavailable, the job can remain pending in the queue.
Earlier integration methods had different constraints. Generic resources can tie a QPU to an individual compute node, while license-based scheduling can represent clusterwide capacity but offers a coarse model. HRES can associate a shared QPU with the nodes that can reach it, so a job can wait for the device without occupying execution nodes during the wait.
Open-source projects built around the Quantum Resource Management Interface (QRMI) and Quantum Device Management Interface (QDMI) use the new HRES interface.
The integrations extend TokenPost’s earlier coverage of hybrid computing software work.