Dimitris Loupas Dimitrios Tsourounis Kostas Blekos Aggelos Tsikas Dimitris Plexousakis Kostas Magoutis Yannis Tzitzikas Haridimos Kondylakis [email protected] [email protected] [email protected] [email protected] Quantum Neural Technologies SA Athens, Greece [email protected] [email protected] [email protected] [email protected] University of Crete, FORTH-ICS Heraklion, Greece Nayak et al. [8] encode bushy join ordering as a QUBO and solve it on quantum annealers/simulators, demonstrating feasibility for small instances. Their work exposes both the potential of quantum-inspired JOO and limitations from variable blow-up and noisy cardinalities, but evaluates only tree validity and optimality—not execution on a real DBMS. Our demo extends this line with a richer, statistics-driven cost model and a join-graph pruning layer that reduces dimensionality before solving, improving scalability and interpretability. A related hybrid strategy is proposed in [6], combining structural encodings with staged decomposition and adaptive solver selection. However, it omits multi-factor weight synthesis from live catalog statistics and lacks connectivity-based subset elimination before binary formulation. In contrast, we pair semantic statistics enrichment with joingraph pruning prior to QUBO construction, producing a leaner variable set and a cost surface closer to classical optimization dynamics. Variational Quantum Circuit (VQC) methods [14] instead learn join orders from historical workloads via hybrid training. Although compact and generalizable, they abstract away multifactor cost semantics and skip explicit pruning of disconnected subsets. Our formulation is complementary: it preserves structural transparency (subset connectivity, explicit penalties) and a tunable cost model while remaining solver-agnostic—enabling future integration of learned proposal distributions to guide search. In this paper we demonstrate Dedalus, a practical bridge between database internals and quantum-enhanced JOO. Our key idea is to transform multi-way join enumeration into a QUBO formulation enriched by: (a) a multi-factor cost model sourced from live database statistics, and (b) an aggressive pruning strategy based on the query’s inferred join graph. The result is a reduced QUBO instance, fed into various (real or simulated) quantum solvers. Compared to combinatorial variable generation, pruning shrinks problem dimensionality, improving solver tractability, stability, and accuracy, leading to better query execution plans, identified faster. Crucially, our work provides an end-to-end implementation that bridges the gap between abstract problem formulations and practical application, demonstrating a complete pipeline from SQL query to solver execution against a live database. To the best of our knowledge, Dedalus is the first end-to-end system that enables the execution and optimization of arbitrary SQL queries over a live database through a fully integrated QUBO-based quantum–classical pipeline. Join order optimization remains one of the most critical and difficult tasks in relational database systems. Even modern cost-based optimizers struggle with exponentially large search spaces and error-prone cardinality estimates, often leading to suboptimal plans that can degrade performance by orders of magnitude. As analytical workloads grow in complexity and scale, there is increasing interest in exploring quantum and quantum-inspired optimization methods to tackle these inherently combinatorial problems. In this demonstration, we present Dedalus, a quantumenhanced framework that formulates join ordering as a Quadratic Unconstrained Binary Optimization (QUBO) problem enriched with database statistics and structural constraints. Dedalus introduces a multi-factor cost model that yields more robust subset weights and a join-graph–guided pruning strategy that eliminates structurally invalid subsets, substantially reducing the dimensionality of the QUBO instance. The system integrates classical simulated annealing, gate-based quantum simulators, exact solvers, and support for various quantum backends, within a single end-to-end pipeline. The demonstration allows users to load SQL queries, inspect the pruned search space, analyze solver-produced join trees, and compare them with PostgreSQL’s native plans, showing how quantum-inspired optimization can already offer practical benefits for join ordering.