Hybrid Quantum-Classical Optimization for Bushy Join Trees
Hanwen Liu, Abhishek Kumar, Federico M. Spedalieri, Ibrahim Sabek · 2025
Optimal join order sequence significantly impacts query execution performance.Because the search space grows exponentially with the number of relations involved in the queries, DP-based approaches can compute optimal plans only for queries with a small number of relations.Many heuristic methods trade off solution quality for computational efficiency.Quantum computing excels at solving combinatorial optimization problems such as join order optimization.However, existing quantum-related methods either focus solely on left-deep join trees-thus narrowing the problem scope-or support bushy join trees but are constrained by quantum hardware and cannot scale to large queries.Importantly, neither branch has been integrated into an actual database system.In this paper, we present a hybrid quantum-classical optimization approach for bushy join trees, integrated within a real database system (PostgreSQL).This approach expands the search space for join order sequences and overcomes previous limitations.Evaluation on the Join Order Benchmark shows that our method can reduce query execution time by up to 92.7% and achieve a 1.42x improvement in end-to-end latency.