🤖 AI Summary
Traditional query optimizers rely on error-prone cost models, while learned optimizers require costly offline training and lack deployment flexibility. Method: We propose a training-free, LLM-embedding-driven execution plan optimization framework. It leverages pre-trained large language models to map SQL queries and their execution plans into a unified embedding space; retrieves semantically similar historical queries via k-nearest-neighbor search; recommends efficient hint sets through majority voting; and ensures robustness via lightweight consistency checking and full-plan-space fallback. Contribution/Results: This is the first work to employ LLM-derived execution plan embeddings for training-free query optimization. It enables end-to-end automatic hint recommendation and achieves deep integration with OpenGauss. Evaluated on the JOB-CEB benchmark, our approach reduces average query latency by 21%, significantly enhancing both practicality and deployment efficiency.
📝 Abstract
Large language model (LLM) embeddings offer a promising new avenue for database query optimization. In this paper, we explore how pre-trained execution plan embeddings can guide SQL query execution without the need for additional model training. We introduce LLM-PM (LLM-based Plan Mapping), a framework that embeds the default execution plan of a query, finds its k nearest neighbors among previously executed plans, and recommends database hintsets based on neighborhood voting. A lightweight consistency check validates the selected hint, while a fallback mechanism searches the full hint space when needed. Evaluated on the JOB-CEB benchmark using OpenGauss, LLM-PM achieves an average speed-up of 21% query latency reduction. This work highlights the potential of LLM-powered embeddings to deliver practical improvements in query performance and opens new directions for training-free, embedding-based optimizer guidance systems.