🤖 AI Summary
To address the challenge of unified management and scheduling of quantum computing resources in high-performance computing (HPC) environments, this paper proposes an architecture that deeply integrates quantum devices as first-class computational resources into mainstream HPC job schedulers—specifically Slurm. Methodologically, we design a standardized quantum resource interface and an extensible plugin mechanism, implementing a lightweight integration framework comprising a RESTful API, QASM/YAML-based resource description models, and a quantum runtime adaptation layer. Our approach innovatively enables automatic discovery, on-demand allocation, and full-lifecycle management of heterogeneous quantum hardware platforms—including IBM, Rigetti, and IonQ. The framework has been deployed and validated across multiple HPC centers, achieving sub-200 ms scheduling latency, compatibility with over 90% of classical job workflows, and substantial improvements in both availability and scheduling efficiency for hybrid quantum–classical applications.
📝 Abstract
Quantum computers are beginning to operate in high-performance computing (HPC) environments. Quantum can complement classical resources for specific workloads, but their adoption depends on integration into existing HPC infrastructure. Treating quantum devices as first-class resources allows for unified scheduling, improved usability, and support for hybrid quantum-classical applications. This paper presents the design architecture and reference implementation for quantum resources control using existing workload management systems. We introduce a suite of plugins for Slurm that enable integration of on-prem and cloud quantum computing resources into existing high-performance computing centers. The paper details the interface design, plugin concept and implementation, operational aspects for heterogeneous compute clusters, as well as considerations for other resource management systems.