DynQ: A Dynamic Topology-Agnostic Quantum Virtual Machine via Quality-Weighted Community Detection
Wednesday, June 24, 2026 3:45 PM to 5:15 PM · 1 hr. 30 min. (Europe/Berlin)
Foyer D-G - 2nd Floor
Research Poster
Integration of Quantum Computing and HPCPerformance and Resource ModelingPerformance Tools and SimulatorsQuantum Program Development and OptimizationQuantum Computing - Technologies and Architectures
Information
Poster is on display and will be presented at the poster pitch session.
Quantum computing is increasingly accessed through cloud and HPC-integrated platforms, yet current quantum systems lack effective virtualisation mechanisms comparable to those used in classical high-performance computing. In today’s quantum clouds, user programs often occupy an entire quantum processing unit (QPU) or large contiguous subgraphs, limiting multi-tenant sharing, throughput, and quality-of-service (QoS) differentiation. This challenge is exacerbated by the heterogeneous and dynamic nature of noisy intermediate-scale quantum (NISQ) hardware, where gate quality varies across the chip and calibration drift, or transient defects can rapidly change hardware usability.
We present DynQ, a dynamic, topology-agnostic quantum virtual machine (QVM) designed to enable robust subgraph virtualisation and multi-program batching on NISQ devices. DynQ is based on the observation that hardware errors are not spatially independent: reliable qubits and couplers tend to form coherent regions, while weaker links concentrate at region boundaries. DynQ exploits this structure by constructing quality-weighted hardware graphs from calibration data and applying community detection to identify high-cohesion, low-coupling regions without relying on handcrafted, topology-specific heuristics.
A key feature of DynQ is its separation of offline region discovery from online allocation. Candidate regions are discovered and scored offline, forming a validated pool of atomic regions. At runtime, DynQ performs low-latency allocation by matching program size and quality requirements to available regions. This design supports practical scheduling, stable multi-program batching, and rapid adaptation to transient hardware degradation. When couplers fail or degrade, DynQ reallocates execution to alternative regions, avoiding catastrophic failures common with static partitions.
We evaluate DynQ using benchmark circuits across multiple simulated IBM backends and real quantum devices. Compared to standard compilation baselines, DynQ improves circuit fidelity by up to 19.1% and reduces total variation distance by up to 45.1% on high-variance hardware. DynQ consistently improves worst-case performance, recovers failed executions, and maintains stable fidelity under increasing batch sizes, while significantly improving effective system throughput.
DynQ demonstrates that topology-agnostic, quality-aware virtualisation is a practical and effective approach for improving reliability, utilisation, and scalability in quantum cloud and HPC-oriented quantum systems.
Contributors:
Quantum computing is increasingly accessed through cloud and HPC-integrated platforms, yet current quantum systems lack effective virtualisation mechanisms comparable to those used in classical high-performance computing. In today’s quantum clouds, user programs often occupy an entire quantum processing unit (QPU) or large contiguous subgraphs, limiting multi-tenant sharing, throughput, and quality-of-service (QoS) differentiation. This challenge is exacerbated by the heterogeneous and dynamic nature of noisy intermediate-scale quantum (NISQ) hardware, where gate quality varies across the chip and calibration drift, or transient defects can rapidly change hardware usability.
We present DynQ, a dynamic, topology-agnostic quantum virtual machine (QVM) designed to enable robust subgraph virtualisation and multi-program batching on NISQ devices. DynQ is based on the observation that hardware errors are not spatially independent: reliable qubits and couplers tend to form coherent regions, while weaker links concentrate at region boundaries. DynQ exploits this structure by constructing quality-weighted hardware graphs from calibration data and applying community detection to identify high-cohesion, low-coupling regions without relying on handcrafted, topology-specific heuristics.
A key feature of DynQ is its separation of offline region discovery from online allocation. Candidate regions are discovered and scored offline, forming a validated pool of atomic regions. At runtime, DynQ performs low-latency allocation by matching program size and quality requirements to available regions. This design supports practical scheduling, stable multi-program batching, and rapid adaptation to transient hardware degradation. When couplers fail or degrade, DynQ reallocates execution to alternative regions, avoiding catastrophic failures common with static partitions.
We evaluate DynQ using benchmark circuits across multiple simulated IBM backends and real quantum devices. Compared to standard compilation baselines, DynQ improves circuit fidelity by up to 19.1% and reduces total variation distance by up to 45.1% on high-variance hardware. DynQ consistently improves worst-case performance, recovers failed executions, and maintains stable fidelity under increasing batch sizes, while significantly improving effective system throughput.
DynQ demonstrates that topology-agnostic, quality-aware virtualisation is a practical and effective approach for improving reliability, utilisation, and scalability in quantum cloud and HPC-oriented quantum systems.
Contributors:
Format
on-demandon-site
