RELOAD: A Robust and Efficient Learned Query Optimizer for Database Systems

TL;DR

RELOAD is a new learned query optimizer that improves robustness and efficiency, reducing performance regressions and training time compared to existing RL-based methods.

Contribution

It introduces a robust and efficient RL-based query optimizer that minimizes regressions and accelerates convergence to expert-level performance.

Findings

RELOAD achieves up to 2.4x higher robustness.

RELOAD is up to 3.1x more efficient in training.

It outperforms state-of-the-art RL-based optimizers on standard benchmarks.

Abstract

Recent advances in query optimization have shifted from traditional rule-based and cost-based techniques towards machine learning-driven approaches. Among these, reinforcement learning (RL) has attracted significant attention due to its ability to optimize long-term performance by learning policies over query planning. However, existing RL-based query optimizers often exhibit unstable performance at the level of individual queries, including severe performance regressions, and require prolonged training to reach the plan quality of expert, cost-based optimizers. These shortcomings make learned query optimizers difficult to deploy in practice and remain a major barrier to their adoption in production database systems. To address these challenges, we present RELOAD, a robust and efficient learned query optimizer for database systems. RELOAD focuses on (i) robustness, by minimizing query-level performance regressions and ensuring consistent optimization behavior across executions, and (ii) efficiency, by accelerating convergence to expert-level plan quality. Through extensive experiments on standard benchmarks, including Join Order Benchmark, TPC-DS, and Star Schema Benchmark, RELOAD demonstrates up to 2.4x higher robustness and 3.1x greater efficiency compared to state-of-the-art RL-based query optimization techniques.