Quantum Computing Is an Orchestration Problem: I Built a Multi-Provider Scheduler and Measured What the Textbooks Only Describe

Описание и применение

Квантовые облачные сервисы страдают от непредсказуемых очередей и неинформативных метрик. Созданный инструмент Quantum Orchestrator автоматически выбирает оптимальный бекенд, измеряет время ожидания и выполнения, а также точность, показывая, что очередь доминирует, а выбор устройства влияет на точность.

Задача

Quantum computing services expose long, unpredictable queues and unreliable backend availability, making scheduling and fidelity selection difficult for users.

Результат

Measured queue-to-execution ratio of 1822:1 on IBM hardware; autonomous backend selection improved fidelity from ~94% to ~97%; a two-basis VQE achieved chemical accuracy, separating shot noise from bias.

Рабочий процесс

• 1. Submit a quantum circuit to Quantum Orchestrator.
• 2. Orchestrator queries each backend adapter for availability and estimated queue.
• 3. It selects the backend with the shortest expected wait and highest expected fidelity.
• 4. The job is submitted; the orchestrator waits, respecting a dynamic fallback threshold and absolute timeout.
• 5. Upon completion, the orchestrator records queue time, execution time, and fidelity.
• 6. Results are reported to the user, including any fallback to simulators if real hardware is unavailable.

Доказательства

On the same IBM quantum processor, on the same afternoon, the same two-qubit circuit took 2 seconds to run and 61 minutes to reach the front of the queue — a queue-to-execution ratio of 1822:1.

Источник

Autonomous backend selection moves not only wait time but fidelity (~94% vs 97% on the same circuit, different processors).

Источник