QRMI expands so supercomputers can book quantum processors like GPUs

Started by TinyCompass, Today at 06:08 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: QRMI expands so supercomputers can book quantum processors like GPUs   Views(Read 42 times)
Active members in this topic:
TinyCompass(1)

TinyCompass

A group of seven organisations has expanded QRMI, the Quantum Resource Management Interface, Quantum Computing Report says. QRMI is open source middleware that lets supercomputer job schedulers reserve quantum processors alongside CPUs and GPUs. The members are IBM Quantum, Pasqal, Alice and Bob, CINECA, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory and the STFC Hartree Centre. The paper is on arXiv as 2607.19591

QRMI now supports six schedulers. They are Slurm through SPANK plugins, PBS Professional, IBM Spectrum LSF, Open Cluster Scheduler, Kubernetes through custom resources and an operator, and Flux. That covers most of the systems used in big computing centres. Researchers will not have to change their usual workflow to use a quantum device

Every integration follows the same three steps. A job reserves the quantum hardware when it starts, receives device details and session tokens while it runs, and releases the reservation when it finishes, whether it succeeds or fails. That last part prevents orphaned sessions and idle holds in the queue. QCR says QRMI decouples application logic and schedulers from hardware APIs through a single middleware layer

There are already real deployments. CINECA links Pasqal's Sol neutral atom processor to its Leonardo supercomputer, a EuroHPC Tier 0 system. RPI connects an IBM Quantum System One to its AiMOS supercomputer, and the BasQ project in Spain is integrating an IBM Quantum System Two. In the UK, the Hartree Centre and the National Quantum Computing Centre are testing it on supercomputer testbeds

This is unglamorous plumbing, but it is exactly what hybrid quantum computing needs. We talked about Japan's JHPC-quantum project last week, and this is a similar idea with an open, vendor neutral approach. If scientists can book quantum time as easily as GPU time, far more of them will try it. Good to see UK centres involved too

Is hybrid quantum and supercomputing the future of practical quantum work? And how important is vendor neutral software for the field?

The truth is usually more complicated than the headline