An Efficient Query Optimization Framework Based on MCTS and LTR

Chaofan Zhao, Yize Sui, Ruochun Jin, Jing Ren

Keywords: Query Optimization, MCTS, Learning-to-Rank

Verify Author List: I have double-checked the author list and understand that additions and removals will not be allowed after the submission deadline.

Abstract: Identifying the optimal query plan is always a fundamental task of query optimization in database management system (DBMS). However, traditional query optimization methods face significant challenges in continuously enhancing query performance due to complex query sentences, intricate data distributions and the exponentially growing search space of table joins. In this paper, we propose a formidable query optimization framework called MRQO (Integrating-MCTS-and-LTR-for-Query-Optimization). This framework utilizes the Monte Carlo Tree Search (MCTS) algorithm to find a comprehensive set of join orders for a query, and uses these join orders as hints to generate corresponding query plans. Additionally, it employs the Learning-to-Rank (LTR) approach to train a relative ranking model, achieving higher efficiency and accuracy in identifying the optimal query plan from all plans. Experimental results on PostgreSQL demonstrate that the proposed MRQO can achieve stable performance and match or even outperform both traditional query optimizers and advanced learned optimizers based on Deep Reinforcement Learning (DRL) in terms of query optimization efficiency.

Primary Area: Applications (bioinformatics, biomedical informatics, climate science, collaborative filtering, computer vision, healthcare, human activity recognition, information retrieval, natural language processing, social networks, etc.)

Paper Checklist Guidelines: I certify that all co-authors of this work have read and commit to adhering to the guidelines in Call for Papers.

Student Author: Yes

Submission Number: 350