Researchers from IBM, Alice & Bob, and Lawrence Livermore National Laboratory have successfully tested the Quantum Resource Management Interface (QRMI) across six distinct workload managers: PBS, LSF, Grid Engine, Kubernetes, Flux Framework and Slurm. This validation extends beyond initial work with Slurm, demonstrating QRMI’s broad compatibility with both traditional batch schedulers and cloud-native orchestration platforms.

QRMI functions as a “thin, vendor-agnostic middleware layer” that exposes quantum resources as first-class schedulable resources alongside CPUs and GPUs, simplifying integration within larger HPC workflows. The work demonstrates that QRMI provides a portable abstraction layer minimizing scheduler-specific modifications while enabling consistent access to quantum resources across diverse environments.

QRMI as a Unified Interface for Quantum-HPC Integration

QRMI’s architecture is designed to encapsulate complexity into configuration and middleware, allowing for a standardized approach to quantum resource allocation within existing HPC infrastructures. This abstraction is achieved through APIs that enable scheduling, execution, and monitoring of quantum workloads, effectively decoupling application logic from the specifics of the underlying quantum hardware. The design prioritizes portability, aiming to present quantum computers as first-class schedulable resources alongside conventional CPUs and GPUs.

This integration is not merely about technical compatibility; it’s about treating quantum processing as a standard component in larger, more complex computational workflows. This broad compatibility is notable given the heterogeneity of these systems, ranging from traditional batch schedulers to cloud-native orchestration platforms like Kubernetes. The approach contrasts with other quantum-HPC integration efforts, such as QDMI (Quantum Device Management Interface), which focuses on compiler toolchain integration and monitoring.

While QDMI and QRMI share some similarities, including a Slurm integration, QRMI specifically targets HPC environments through workload managers, offering a complementary solution. The work demonstrates a shift towards treating quantum resources as integral parts of HPC systems, rather than isolated entities requiring bespoke management solutions.

QRMI Architecture and Design for Heterogeneous Resources

QRMI’s architecture deliberately decouples workload manager logic from the specific application programming interfaces of quantum hardware, enabling a consistent operational workflow across multiple quantum providers and deployment models. This abstraction allows quantum resources to be represented uniformly, regardless of whether they are cloud-hosted, a single quantum processing unit, a device partition, or parallel execution lanes, simplifying administration and unifying scheduling, accounting and access control. The design, detailed in the paper, uses core language bindings and resource-management functions alongside adapters connecting workload managers and quantum software development kits to diverse quantum resources.

These traditional schedulers demonstrated partial support for treating quantum resources as schedulable entities and encapsulating complexity into configuration and middleware, but lacked native quantum resource support. Kubernetes and Flux Framework, representing cloud-native orchestration, were included in the broader validation of QRMI’s compatibility.

The findings from these integrations are feeding back into the openQSE specification effort, creating a validation loop where architectural decisions are tested in a working system before being proposed as community standards. The success of QRMI in supporting these design patterns is significant because it addresses concrete challenges in quantum-HPC integration. The paper highlights that QRMI’s approach allows users to submit hybrid quantum-HPC jobs through familiar workload manager interfaces, which is an important step towards wider adoption.

“QRMI serves as a vendor-agnostic abstraction layer that enables quantum computers to be managed as first-class schedulable resources within existing HPC workload management systems,” the authors state, emphasizing the goal of smooth integration. This decoupling is intended to minimize the need for hardware-specific modifications, allowing the same workflow to support diverse quantum hardware and deployment scenarios, and ultimately, simplifying the administration of increasingly complex quantum-enabled HPC environments.

QRMI Integration with Slurm Workload Manager

The design of QRMI, using thin integration layers, aims to mitigate differences and provide a consistent interface for quantum resource management. While six workload managers, Slurm, PBS, LSF, Grid Engine, Kubernetes and the Flux Framework, permit mapping lifecycle events to QRMI operations via plugins or hooks, not all fully support the creation of those thin integration layers. The level of support for key design patterns across these systems is summarized in Table II.

The assessment demonstrates that encapsulating complexity within configuration and middleware is broadly achievable, but native quantum resource support remains absent from all tested platforms. This finding highlights the current reliance on abstraction layers like QRMI to bridge the gap between existing HPC infrastructure and emerging quantum technologies. The paper highlights that QRMI’s approach to lifecycle management, using plugins, is repeatable but not universally portable due to inherent differences in workload manager design.

Validating QRMI Across Traditional Batch Schedulers

This broad assessment encompassed both traditional batch scheduling systems and more modern approaches like cloud-native orchestration and graph-based scheduling, revealing the interface’s adaptability across diverse computing environments. This validation is notable because it extends QRMI’s reach into systems not originally designed to handle the unique demands of quantum computing. Traditional schedulers often struggle with the latency constraints, session requirements, and limited availability windows inherent in quantum hardware, frequently requiring manual intervention for workflow orchestration and data movement.

QRMI aims to alleviate these issues by providing a consistent abstraction layer, simplifying administration and enabling tighter coordination between classical and quantum resources. The team’s analysis focused on integration patterns, implementation requirements and scheduler-specific considerations for each environment, identifying areas for improvement in both QRMI and the workload managers themselves.

The evaluation also highlighted the importance of resource discovery and credential management. Quantum technologies expose diverse capabilities, while schedulers employ varying resource models and selection mechanisms. Current QRMI prototypes use metadata to represent these properties, with future development planned to define simpler, more compact abstractions that can be mapped to scheduler-specific models. Addressing credential handling is important for secure access, with options including administrator-managed files and environment variables currently under consideration. “Resource discovery is challenging because quantum technologies expose different capabilities,” the researchers note, emphasizing the need for standardized representations of quantum resource characteristics.

Quantum Resource Management Challenges in HPC Environments

While all tested systems offer extension mechanisms for scheduling and resource management, dynamic quantum resource availability presented consistent challenges. Provider-managed queues, calibration downtime, and opaque execution latencies can all diminish overall system use without additional scheduling logic or resource awareness. These observations suggest that future workload managers will require enhanced support for dynamic and heterogeneous resources beyond current HPC norms. The research also highlights limitations inherent in existing workload managers when applied to quantum computing.

The team found that while all systems provide mechanisms for extending scheduling and lifecycle management, none offer native support for quantum resources. This lack of native support necessitates workarounds and potentially compromises efficiency. The findings emphasise a growing demand for workload managers and orchestration systems capable of handling the unique demands of quantum-HPC environments.

👉 More information
🗞 Examining QRMI as a Unified Interface for Quantum-HPC Integration
✍️ Thomas Badts et al.
🧠 ArXiv: https://arxiv.org/abs/2607.19591

More like this