logo
CanyamAI summaries for academic research
language
MobileMobile
  • Phone FeatureWeChat Mini Program
EnglishEnglish
noticeNotice
BlogBlog
delete
arrow
Return

Verified LLM-Based Query Rewriting for Microsoft SQL Server

delete2026-08-01
delete0
PRE
AI
K
Kukjin Lee
A
Anshuman Dutt
V
Vivek Narasayya
S
Surajit Chaudhuri
DOI：10.14778/3827998.3828074delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
Query rewriting is one of the techniques used by database developers and administrators to tune the performance of poorly performing SQL queries. We have developed an LLM-based query tuning tool for Microsoft SQL Server that takes as input a query, tuning constraints such as minimum required improvement and time budget, and automatically recommends semantically equivalent query rewrites that improve the execution time of the query. The software tool is incorporated as an extension to SQL Server Management Studio (SSMS). Our demonstration highlights two important capabilities. First, while LLMs have the ability to generate promising candidate rewrites for the given query, they cannot guarantee semantic equivalence of a query rewrite. We use a technique we developed for verifying semantic equivalence of two queries, which leverages the built-in capabilities of Microsoft SQL Server's query optimizer. Second, we judiciously use the given time budget by adopting an anytime tuning algorithm and employing optimizations that help short-circuit the expensive steps of verification and execution for each candidate rewrite. To the best of our knowledge this is the first tool of its kind that enables DBAs and application developers to safely operationalize LLM-based query rewriting.

Journal

P
Proceedings of the VLDB Endowment
IF:
3.3
Papers:
563
Citations:
1.2W

Organization

No organization information available