The access to quantum hardware is limited, which is why simulating is common practice in quantum computing research. Thanks to granted access to the QExa20 quantum computer at the Leibniz Supercomputing Centre during its pilot phase, this thesis focuses on reproducing previously simulated experiments on real quantum hardware. The QExa20 is part of the European Quantum Computing for Exascale-HPC project and the first of its kind connected with a SuperMUC high-performance computer. There are five more quantum computer systems planned across Europe, with the vision of making several quantum computers and simulators accessible for multiple users.
The experiments ported to the QExa20 are based on the approaches introduced by Krรผger and Mauerer (2024) and Franz et al. (2024), presenting a non-iterative version of the Quantum Approximate Optimization Algorithm (QAOA) and a quantum reinforcement learning method for solving the join order problem. Although the system is still in an early stage of development, it faces initial challenges, such as long runtimes and unclear error responses. Because 20 qubits is a really limited source to work with, the experimental parameters needed to be scaled down to fit the number of available qubits and experiments were carefully selected to account for long execution times.
As far as this thesis examined the performance of the QExa20, a really low depolarization error could be obtained and the results for non-iterative QAOA could be confirmed on real quantum hardware. Despite all limitations, the QExa20 has proven to be capable of producing reliable results. However, integrating quantum hardware into large computer systems and controlling multi-user access remains a complex challenge. Network latencies and request management can easily erode quantum advantage.