Dynq Achieves Virtualisation of Quantum Hardware Using Quality-Weighted Community Detection

Researchers are tackling a critical limitation in quantum computing: the inability to truly virtualise quantum hardware. Shusen Liu, Pascal Jahan Elahi, and Ugo Varetto, all from the Pawsey Supercomputing Research Centre at the University of Western Australia, introduce DynQ, a dynamic and topology-agnostic quantum virtual machine that overcomes the constraints of existing, brittle approaches to resource sharing. Unlike current Virtual Machine designs which rely on fixed hardware partitioning, DynQ utilises quality-weighted community detection to model a processor as a weighted graph, automatically discovering execution regions optimised for gate quality and resilience, effectively bringing the principles of classical virtualisation to the quantum realm. This innovation, demonstrated across IBM backends, achieves significant gains in fidelity (up to 19.1 percent) and error reduction (45.1 percent), and crucially, allows for continued operation even when hardware defects cause traditional executions to fail, paving the way for reliable and scalable quantum cloud services.

The research addresses the fundamental limitation of current cloud platforms where each user program monopolises an entire quantum processor, hindering resource sharing and economic scalability. Instead of relying on fixed geometric regions for virtualisation, the team modelled a quantum processor as a weighted graph derived from live calibration data, automatically discovering execution regions that maximise internal gate quality while minimising inter-region coupling. This innovative approach operationalises the classical virtualisation principle of high cohesion and low coupling within a quantum-native setting, resulting in execution regions that are connectivity-efficient, noise-aware, and resilient to Crosstalk and defects.

Experiments conducted across five IBM Quantum backends, utilising calibration-derived noise simulations and on two production devices, showcase DynQ’s performance compared to state-of-the-art QVMs and standard compilation baselines. On hardware exhibiting pronounced spatial quality variation, DynQ achieved up to 19.1 percent higher fidelity and a 45.1 percent reduction in output error, demonstrating a substantial improvement in computational accuracy. Crucially, the study reveals DynQ’s ‘dead-link immunity’; when transient hardware defects caused baseline executions to fail, DynQ dynamically adapted and maintained over 86 percent fidelity, highlighting its robustness and resilience. This dynamic adaptation is achieved by transforming calibrated device graphs into adaptive virtual hardware abstractions, effectively decoupling quantum programs from fragile physical layouts.
The research establishes that DynQ’s ability to dynamically adjust to hardware conditions enables reliable, high-utilisation cloud services and addresses the critical need for resource sharing in quantum computing. Under multi-tenant execution, fidelity remained statistically stable (r = −0.012) as concurrency scaled from 2 to 18 programs, allowing for up to a 90 percent reduction in job cost through safe batching. By decoupling programs from physical layouts, DynQ overcomes limitations of existing QVMs that rely on static mappings and struggle with hardware heterogeneity, calibration drift, and transient defects. This breakthrough establishes a principled foundation for topology-independent, quality-aware quantum virtualisation as quantum processors scale towards thousands of qubits and increasing architectural complexity.

Dynamic Quantum Virtualisation via Weighted Community Detection offers

Scientists pioneered DynQ, a dynamic, topology-agnostic Virtual Machine (QVM) that virtualises quantum hardware using quality-weighted community detection. Instead of relying on fixed geometric regions, the research team modelled processors as weighted graphs derived from live calibration data, automatically discovering execution regions that maximise internal gate quality and minimise inter-region coupling. This approach operationalises the principle of high cohesion and low coupling within a native quantum setting, creating execution regions that are connectivity-efficient, noise-aware, and resilient to crosstalk and defects. The study employed an algorithm that operates on any coupling graph without modification, dynamically re-discovering regions to handle transient defects, a significant advancement in quantum virtualisation.
Experiments involved evaluating DynQ across five simulated IBM backends and two production quantum devices, comparing its performance against state-of-the-art QVMs and standard compilation baselines. Researchers meticulously measured calibration-derived noise to inform the weighting of the processor graphs, ensuring accurate representation of hardware quality. The team then implemented a modularity optimisation technique, formalising quantum resource partitioning to produce regions with strong internal connectivity and weak external coupling, directly linking quantum virtualisation to established graph algorithms. This partitioning methodology enabled the creation of robust crosstalk suppression under realistic multi-tenant concurrency, demonstrated by maintaining statistically stable circuit fidelity (r = −0.012) as the number of concurrent tenants increased from 2 to 18.

Furthermore, the study showcased DynQ’s fault recovery capability, achieving over 86 percent fidelity when standard compilation resulted in complete failures due to transient hardware defects. This was accomplished by dynamically adapting to hardware issues and re-allocating resources, a crucial step towards reliable quantum cloud services. On hardware exhibiting pronounced spatial quality variation, DynQ achieved up to 19.1 percent higher fidelity and a 45.1 percent lower output error, quantified by a reduction in total variation distance to the ideal output distribution. Batch scalability experiments confirmed stable multi-tenant performance, demonstrating the potential for substantial cost reduction through job batching and efficient resource utilisation.

DynQ boosts fidelity via quality-weighted virtualisation, enhancing realism

Scientists have developed DynQ, a dynamic, topology-agnostic Virtual Machine that virtualises quantum hardware using quality-weighted community detection. The team measured up to 19.1 percent higher fidelity and 45.1 percent lower output error on hardware exhibiting pronounced spatial quality variation, demonstrating a significant leap in performance. Experiments revealed that DynQ models a processor as a weighted graph derived from live calibration data, automatically discovering execution regions that maximise internal gate quality while minimising inter-region coupling. This innovative approach operationalises the principle of high cohesion and low coupling in a native setting, creating execution regions that are connectivity-efficient, noise-aware, and resilient to crosstalk and defects.

Results demonstrate that DynQ achieves over 86 percent fidelity in scenarios where standard compilation results in complete failures due to transient hardware defects, a remarkable recovery capability. The research team measured circuit fidelity remaining statistically stable (r = −0.012) as the number of concurrent tenants increased from 2 to 18, indicating that inter-tenant interference does not accumulate with load. This stability confirms that quality-weighted community boundaries provide robust crosstalk suppression under realistic multi-tenant concurrency, paving the way for efficient resource sharing. By assigning lower weights to low-fidelity couplers, DynQ’s modularity objective naturally places region boundaries along interference-prone links, effectively creating implicit buffer zones without reserving dedicated idle qubits.

Tests prove that DynQ decouples programs from fragile physical layouts, enabling reliable, high-utilisation cloud services. The algorithm operates on any coupling graph without modification, dynamically rediscovering QVM regions to handle transient defects. Measurements confirm a 45.1 percent reduction in the total variation distance to the ideal output distribution, corresponding to the 19.1 percent improvement in result fidelity on high-variance hardware. This breakthrough delivers a fault recovery capability, allowing program execution with over 86 percent fidelity when standard compilation fails completely.

Furthermore, the work formalises quantum resource partitioning as modularity optimisation, producing regions with strong internal connectivity and weak external coupling, essential for efficient circuit compilation. Extensive experiments were conducted across five simulated IBM backends and two real quantum devices, comparing DynQ against standard Qiskit compilation as a baseline. Batch scalability experiments demonstrate stable multi-tenant performance under increasing concurrency, enabling substantial cost reduction through job batching and highlighting the potential for scalable quantum cloud services.

DynQ enables dynamic quantum resource virtualisation for enhanced

Scientists have developed DynQ, a dynamic and topology-agnostic Quantum Virtual Machine that reimagines quantum resource virtualisation. Unlike previous systems requiring manual configuration for specific hardware, DynQ automatically discovers virtual machine regions using quality-weighted community detection. This innovative approach achieves topology independence, dynamic adaptation to hardware defects, and quality differentiation for improved service allocation. The core methodological advancement lies in framing quantum resource partitioning as modularity optimisation on quality-weighted graphs, effectively applying the systems principle of high cohesion and low coupling to quantum computing.

By maximising internal connectivity within virtual machine regions and minimising coupling between them, DynQ suppresses interference and enables reliable operation. Evaluation across five simulated backends and two real quantum devices demonstrates significant improvements in fidelity and error reduction, particularly on hardware with substantial quality variations. Notably, DynQ maintained over 86% fidelity even when baseline executions failed completely due to transient defects, showcasing its resilience. The authors acknowledge that the current implementation is evaluated on a limited number of hardware backends and devices. Future research will likely focus on scaling DynQ to larger quantum processors and exploring its performance across a wider range of architectures. However, this work establishes that graph-based approaches offer the flexibility, reliability, and efficiency needed for practical multi-tenant quantum cloud services, paving the way for cost-effective and high-utilisation quantum computing.