Extended Quantum Computing Patterns

How to execute multiple quantum circuits in succession while keeping the queuing time low?

Quantum applications often require executing multiple quantum circuits. This is especially the case when utilizing VQAs on contemporary NISQ devices. Thus, these quantum circuits should be executed efficiently by minimizing the queuing times.

Quantum devices are usually accessed via queues, ensuring their fair utilization. Thus, the queuing times sum up when executing multiple quantum circuits independently of each other. One approach is to use batch processing, i.e., combining various quantum circuits into one job, which is then executed at once [Vietz et al., 2021]. However, this approach is not possible when quantum circuits depend on the results of previous executions, e.g., for VQAs.

Use quantum offerings that enable prioritized access to quantum devices to reduce or completely avoid queuing times, as shown in the solutions sketch in Figure 5. For this, prioritized access to quantum devices is obtained in Step 1. In Step 2, the quantum device can then be reused via this prioritized access, which restricts the number of users to reduce queuing times for multiple circuit executions.

By applying this pattern, the time spent waiting inside highly occupied job queues is minimized. The choice between the two options is a trade-off between runtime and cost. In the reserved time-slice scenario, runtime is the highest priority since the reservation of a quantum device incurs higher costs than other forms of access. When using sessions for the execution, there can still be competing executions in the job queue from other users.

–

When using the VQA, VQE, or QAOA patterns [Weigold et al., 2021b], prioritization is highly advised because of the amount of sequential quantum circuit executions. In these cases, it can be combined with the PRE-DEPLOYED EXECUTION pattern to address custom deployment requirements.

The Session Reuse variant is currently supported by IBM [Qiskit, 2023a] and AWS [AWS, 2023]. Further, the Dedicated Access variant is offered by vendors such as AWS [AWS, 2023], Azure [Microsoft, 2023], or IBM [IBM, 2023b]. A concrete example of a prioritized execution using Qiskit [Qiskit, 2023b] highlights which options for the session are available and shows the details like the session time limit.