A Case for Agentic Tuning: From Documentation to Action in PostgreSQL

2026-05-19

AI Summary

This work addresses the limitations of traditional database parameter tuning, which relies on static documentation and struggles to adapt to version evolution, heterogeneous workloads, and complex inter-parameter dependencies. The authors propose a novel large language model–based approach that transforms expert knowledge into executable tuning skills, enabling version-consistent validation, workload-aware analysis, and joint optimization of multiple parameters. By dynamically operationalizing the tuning process, this method overcomes the constraints of static guidelines and supports adaptive configuration tailored to real-world execution environments. Experimental evaluations on PostgreSQL using TPC-C and TPC-H benchmarks demonstrate performance improvements of up to 35.2% compared to existing documentation-driven tuning strategies.

Abstract

Documentation has long guided computer system tuning by distilling expert knowledge into per-parameter recommendations. Yet such guides capture only what experts conclude, discarding how they reason. This fundamental gap manifests in three concrete deficiencies: documentation grows stale as software evolves, fails under heterogeneous workloads, and ignores inter-parameter dependencies. We propose shifting from static documentation to dynamic action for system tuning. We introduce PerfEvolve, which translates expert tuning methodologies into executable skills that equip LLM-based agents to perform version-consistency verification, workload-specific profiling, and multi-parameter joint optimization. Evaluated on PostgreSQL under TPC-C and TPC-H benchmarks, PerfEvolve outperforms state-of-the-art documentation-driven tuning baselines by up to 35.2%. The tool is available at https://github.com/ISCAS-OSLab/PerfEvolve.

Problem

• system tuning
• documentation limitations
• parameter dependencies
• workload heterogeneity
• version evolution

Innovation

• Agentic Tuning
• PerfEvolve
• LLM-based agents
• multi-parameter optimization
• workload-specific profiling