IDSTune: A Multi-Agent Collaborative Framework for Integrated Database System Tuning

LLM-driven multi-agent collaboration jointly optimizes knobs, indexes, and materialized views for globally consistent DBMS tuning

Yiyan Li; Guanli Liu; Renata Borovica-Gajic; Haoyang Li; Zihang Qiu; Xinmei Huang; Andreas Kipf; Cuiping Li; Hong Chen· 2026· DOI 10.48550/arXiv.2607.22031

The core problem

Database tuning is critical for achieving high performance in modern database management systems (DBMSs). Existing methods typically optimize a single component—knobs, indexes, or materialized views—without accounting for their interdependencies. This limitation arises because these components require different tuning strategies and are difficult to integrate within a unified framework. As a result, directly extending a method to multiple components or simply combining separate methods often fails to capture cross-component collaboration and shared tuning signals. Moreover, existing methods are insufficient for handling diverse workloads, evolving data, and dynamic query patterns. To address these limitations, the authors propose IDSTune, an integrated tuning framework that jointly optimizes multiple configuration components through LLM-driven multi-agent collaboration.

Innovation

Extensive experiments show that IDSTune achieves up to 38% performance improvement and 57% faster tuning, with strong adaptability across diverse scenarios. These results demonstrate the effectiveness of integrated tuning over single-component or naively combined approaches.

Database tuning is critical for achieving high performance in modern database management systems (DBMSs). Existing methods typically optimize a single component—knobs, indexes, or materialized views—without accounting for their interdependencies. This limitation arises because these components require different tuning strategies and are difficult to integrate within a unified framework. As a result, directly extending a method to multiple components or simply combining separate methods often fails to capture cross-component collaboration and shared tuning signals. Moreover, existing methods are insufficient for handling diverse workloads, evolving data, and dynamic query patterns. To address these limitations, the authors propose IDSTune, an integrated tuning framework that jointly optimizes multiple configuration components through LLM-driven multi-agent collaboration.

IDSTune operates in two phases: (i) workload compression, which extracts and selects task-relevant features, and (ii) configuration recommendation, where specialized agents collaboratively generate and refine configurations for knobs, indexes, and materialized views under the supervision of a centralized coordinator. By incorporating feedback and external knowledge retrieval, IDSTune achieves efficient and globally consistent tuning.

Why it matters

The key innovation of IDSTune lies in its ability to capture cross-component collaboration and shared tuning signals through LLM-driven multi-agent collaboration. By jointly optimizing knobs, indexes, and materialized views, it overcomes the limitations of traditional methods that treat these components in isolation. The two-phase design—workload compression followed by configuration recommendation—enables efficient handling of diverse workloads, evolving data, and dynamic query patterns. The centralized coordinator ensures globally consistent tuning, while feedback and external knowledge retrieval allow continuous adaptation. The reported performance gains (up to 38% improvement and 57% faster tuning) highlight the potential of integrated tuning frameworks for modern DBMSs. Future work may explore scaling to more components and integrating additional tuning dimensions.

Who should read this

CS practitioners and researchers

Opening member content…