Fragmented quantum cloud infrastructures now pose a key engineering challenge for developers seeking predictable performance from their circuits. A framework called Quantum Broker distributes individual quantum computations across multiple heterogeneous processors based on user-specified Quality-of-Service (QoS) requirements. The University of Pisa and University of Extremadura team developed the Quantum Broker to manage quantum computing tasks across different machines.

As cloud access to varied quantum processors increases, effectively dividing work between them becomes key; this system intelligently assigns parts of calculations based on qualities like speed or accuracy. This allows for better control over resources when running complex programs on current hardware by optimising performance despite differences in available technology. As cloud access to varied quantum processors increases, managing computations across them presents a key engineering challenge; the University of Pisa and University of Extremadura team have responded with the Quantum Broker framework.

This system intelligently distributes individual calculations based on qualities like speed or accuracy, a process akin to dividing up a large image for rendering by multiple servers, to optimise performance despite differences in available technology. The core idea is shot-wise distribution, breaking down one task into many smaller pieces sent individually to different computers.

An open-source set of tools allows users to specify Quality-of-Service (QoS) requirements, such as cost limits or desired fidelity, then automatically assigns work accordingly using algorithms that balance competing priorities, much like an advanced route planner considers traffic when finding the fastest way home. But can this broker effectively navigate the complexities of fragmented quantum cloud infrastructures and deliver truly predictable results.

Quantum Brokers enable successful circuit completion on heterogeneous processors

Quality of Service (QoS) satisfaction demonstrably improves with the Quantum Broker framework compared to existing methods; experiments reveal measurable gains in achieving user-defined constraints across various quantum processors where single or naive multi-backend approaches previously failed to deliver feasible plans. The system uniquely distributes individual measurement ‘shots’, fundamental units of quantum computation, across multiple QPUs based on parameters like cost, latency, and fidelity using either mixed-integer linear programming or dual annealing solvers. Combining a modular execution architecture with interchangeable backend profiling strategies alongside the new Quantum QoS Specification Language (QQSpec), users can declaratively express requirements for their circuits.

Valid execution plans were generated by the team across diverse quantum backends, successfully dispatching circuits when both single-backend methods and simpler multi-backend techniques fell short, a vital step toward reliable cloud access. Experiments utilising benchmark circuits revealed that this framework improved Quality of Service (QoS) satisfaction compared to baseline systems, achieving feasible solutions where previously none existed.

While the dual-annealing solver produced higher quality allocations than mixed integer linear programming, it came at an increased computational cost for planning; specifically, better shot distribution was achieved with more complex backend profiles. The system’s modular design also allows users to define requirements using QQSpec, enabling declarative expression of circuit needs like latency or fidelity.

Simulations reveal resource management strategies for future quantum cloud deployment

Efficiently leveraging a growing pool of diverse processors is crucial for realising the promise of cloud-based quantum computation; however, simply finding *a* backend isn’t enough, truly useful systems must actively manage resources to meet user demands such as cost or speed. Current evaluation relies heavily upon simulations and benchmark circuits which may not fully capture real-world complexities like queueing delays in actual cloud environments. Current evaluation relies heavily upon simulations and benchmark circuits which may not fully capture real-world complexities like queueing delays.

It’s important to acknowledge that this work currently depends on simulations rather than live cloud deployments, as real-world performance will certainly differ due to factors like queueing delays and system overheads. However, it establishes a key foundation by demonstrating complex calculation management across multiple processors simultaneously; offering a flexible layer for building portable quantum software capable of adapting to fragmented infrastructures while preparing for scale even as commercial platforms mature. The researchers and University of Extremadura developed Quantum Broker, a framework designed to intelligently distribute individual quantum calculations, known as ‘shots’, across diverse processors concurrently. Exploiting the inherent independence of each measurement shot enables execution management over fragmented cloud infrastructures alongside an extensible foundation for portable software development.

The researchers developed Quantum Broker, an open-source framework that distributes individual quantum calculations, called shots, across multiple heterogeneous QPUs simultaneously. This approach allows users to define quality-of-service requirements like latency or fidelity through QQSpec, enabling more control over circuit execution in cloud environments. Experiments using benchmark circuits and simulations demonstrated this system can create feasible dispatch plans under varying backend conditions and improve satisfaction of those defined QoS constraints. The modular design supports interchangeable profiling strategies and two solver backends: mixed-integer linear programming and dual annealing.

👉 More information
🗞 Quantum Broker: QoS-Aware Distribution of Quantum Computations
✍️ Giuseppe Bisicchia, Alessandro Bocci and Antonio Brogi (University of Pisa); Jose Garcia-Alonso and Juan Manuel Murillo (University of Extremadura)
🧠 ArXiv: https://arxiv.org/abs/2610.28342

More like this