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

TL;DR

This paper introduces PerfEvolve, a dynamic agent-based system that automates PostgreSQL tuning by translating expert knowledge into executable skills, outperforming traditional documentation-based methods.

Contribution

It presents PerfEvolve, a novel approach that converts expert tuning strategies into executable actions for LLM agents, addressing limitations of static documentation.

Findings

PerfEvolve outperforms baseline tuning methods by up to 35.2% on PostgreSQL benchmarks.

The system effectively performs version verification, workload profiling, and multi-parameter optimization.

Evaluation on TPC-C and TPC-H benchmarks demonstrates significant performance improvements.

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.