Researchers Distribute Shots across QPUs for Better Fidelity

Distributing Quantum Shots for Better Fidelity

Cloud‑based quantum computing is increasingly fragmented: users can tap a growing pool of heterogeneous processors, each with its own speed, error rates, and pricing. The challenge is not just finding a backend, but orchestrating work so that a user’s quality‑of‑service (QoS) requirements—such as target fidelity, budget, or latency—are reliably met. Researchers at the University of Pisa and the University of Extremadura address this with Quantum Broker, an open‑source framework that breaks a quantum job into its constituent measurement shots and allocates those shots across multiple QPUs.

---

How Shot‑Wise Distribution Works

A quantum circuit is executed many times; each execution (a shot) yields a bitstring that contributes to the final statistical estimate. Because shots are independent, they can be dispatched to different machines without altering the algorithmic outcome. Quantum Broker exploits this independence by:

1. Profiling backends – gathering latency, cost, and fidelity metrics for each available QPU.
2. Accepting a QoS specification – users write declarative constraints in the new Quantum QoS Specification Language (QQSpec), e.g., “average fidelity ≥ 99 % and total cost ≤ $0.10”.
3. Solving an allocation problem – the framework offers two solvers: a mixed‑integer linear programming (MILP) model for exact optimization and a dual‑annealing heuristic that often finds higher‑quality allocations at the expense of longer planning time.
4. Generating an execution plan – the plan maps a number of shots to each selected backend, respecting the QoS constraints.
5. Running the distributed job – shots are sent to the chosen QPUs, results are collected, and the final distribution is reconstructed.

---

Benefits Demonstrated in Simulation

Using benchmark circuits (e.g., variational quantum eigensolver instances and random Clifford circuits), the authors compared Quantum Broker against two baselines:

• Single‑backend selection – picking the best‑rated QPU and sending all shots there.
• Naïve multi‑backend – splitting shots evenly without QoS awareness.

Key findings include:

• Higher QoS satisfaction – the broker produced feasible plans for 87 % of test cases, whereas the single‑backend approach succeeded in only 53 % and the naïve multi‑backend in 61 %.
• Improved fidelity – by routing more shots to higher‑fidelity devices while keeping cost within budget, average circuit fidelity rose by up to 3 percentage points.
• Solver trade‑offs – the dual‑annealing solver achieved the best fidelity improvements but required roughly twice the planning time of the MILP solver.

These results illustrate that intelligent shot distribution can turn a fragmented quantum cloud into a cohesive resource pool, delivering more predictable outcomes for developers.

---

Practical Considerations and Limitations

While the simulation results are promising, several real‑world factors remain to be explored:

• Queueing delays – live cloud environments introduce variable wait times that can affect latency guarantees.
• Backend profiling accuracy – the quality of the allocation depends on up‑to‑date performance data; stale metrics could lead to suboptimal plans.
• Overhead of planning – especially with the dual‑annealing solver, the time spent computing an allocation may offset gains for short‑running jobs.
• Security and data locality – distributing shots across providers may raise compliance concerns for sensitive workloads.

Future work should integrate live queue monitoring, adaptive profiling, and cost‑aware security policies to bridge the gap between simulation and production use.

---

Looking Ahead

Quantum Broker provides a modular foundation for portable quantum software that can adapt to an ever‑more diverse cloud landscape. By exposing a declarative QoS language and interchangeable solvers, it invites extensions such as:

• Dynamic re‑allocation – adjusting shot distribution on‑the‑fly as backend performance changes.
• Hybrid classical‑quantum pipelines – coordinating shot distribution with classical post‑processing workloads.
• Marketplace integration – allowing users to source shots from third‑party providers that expose standardized QoS metadata.

As commercial quantum platforms mature and the number of accessible QPUs grows, frameworks like Quantum Broker will be essential for turning raw hardware variety into reliable, user‑centric cloud services.

---

The Quantum Broker framework and its associated tools are available under an open‑source license, encouraging the community to experiment, contribute, and extend its capabilities.