QQLink: A Declarative Framework for Multi-QPU Quantum Workflow Orchestration
This paper introduces QQLink, a declarative framework that orchestrates multi-QPU quantum workflows by automatically deriving cross-node dependencies from a single program, demonstrating high-fidelity simulation results and successful execution on partitioned IBM Heron hardware while highlighting current limitations in physical networked QPU coherence.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Quantum computers promise to solve problems that are impossible for today's machines, but they face a fundamental size limit. The delicate quantum bits, or qubits, that hold information are difficult to keep stable, and packing too many into a single processor often causes them to lose their state. To build larger, more powerful machines, scientists are looking toward a different approach: connecting many smaller quantum processors together. Imagine a network of these small devices working in unison, where a measurement taken on one machine instantly influences a calculation on another. This is the vision of distributed quantum computing. However, making this work requires a new kind of software. The processors cannot simply be told to run separate tasks; they must coordinate their actions in real time, sharing information and adjusting their behavior based on what their partners discover. Without a way to manage this complex conversation, the network remains just a collection of isolated parts.
Researchers at Oak Ridge National Laboratory have developed a new software framework called QQLink to solve this coordination problem. Instead of asking a programmer to write separate instructions for each quantum processor and then manually figure out how they should talk to each other, QQLink allows a single program to describe the entire workflow. In this system, a developer defines the steps for different parts of the network, known as workflow nodes, and declares how they are connected. The software then automatically figures out the necessary order of operations. It ensures that if one node measures a qubit, the other nodes know exactly when to use that result to adjust their own calculations. The researchers tested this system by running seven different quantum protocols, including methods for sending information and performing complex searches, across simulated environments and on actual quantum hardware.
The core achievement of QQLink is its ability to translate a high-level description of a multi-processor task into a precise, step-by-step plan that the hardware can follow. The researchers found that the software could generate these plans extremely quickly, taking only microseconds to organize even large sets of instructions. When they ran the software on a simulator, the results were extremely precise. They tested every possible outcome of the measurements and found that the final quantum states matched the perfect theoretical results with an infidelity of at most 5 × 10⁻¹², and there were no incorrect classical outputs. This proved that the logic of the framework was sound and that the automatic coordination worked exactly as intended.
To see if this worked on real machines, the team ran the programs on two IBM quantum processors. They did not use two separate physical computers connected by a cable; instead, they treated different sections of a single chip as if they were separate processors. This allowed them to test the software's ability to manage "virtual" quantum processors that share the same physical space. The results showed that the software could successfully direct these virtual processors to work together in terms of classical outcomes. However, the real-world runs also revealed the current limits of the hardware. While the software logic was perfect, the physical qubits were noisy. The success rate of the complex tasks dropped as the number of steps increased, and the results varied noticeably between different runs on the same machine.
The researchers discovered that the hardware struggled with the timing and sensitivity required by these distributed tasks. In some cases, the quantum state did not preserve the delicate superposition needed for the calculation to work, even though the software had correctly ordered the steps. Specifically, coherence-sensitive checks did not establish coherent Toffoli operation or the preservation of the intended repetition-code superposition. This suggests that while the software framework is ready to manage a network of quantum computers, the physical machines themselves are not yet stable enough to fully support the most complex distributed algorithms. The study confirms that the software can derive the correct dependencies and execute them, but it also highlights that the physical realization of these networks faces significant challenges with noise and coherence.
One of the most important findings was that the software could run the same program on a simulator and on the hardware without any changes to the code. This portability is a crucial step toward a future where quantum applications can be written once and run on different types of machines. The researchers also tested a scenario where the different parts of the network ran on separate computer processes, communicating only through delayed messages. Even in this distributed setup, where the parts did not share a central controller, the software ensured the tasks completed correctly. This demonstrates that the framework can handle the complexity of a true network, not just a single chip divided into sections.
Despite the hardware limitations, the work provides a clear path forward. The researchers showed that the coordination logic does not need to be written by hand for every new task; it can be generated automatically from a single description. They also identified that the current hardware struggles with the specific timing of these operations, a problem that might be solved by better hardware or by adjusting how the software sends commands. The study does not claim to have solved the problem of building a global quantum internet, but it has built the essential software layer that would be required to make it work. By proving that a single program can successfully orchestrate multiple quantum processors, QQLink moves the field closer to a future where quantum computers can scale up by working together rather than by simply getting bigger.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.