KnobTree: Intelligent Database Parameter Configuration via Explainable Reinforcement Learning

TL;DR

KnobTree tackles the challenge of tuning database configurations with hundreds of tunable parameters where traditional rule-based methods struggle and deep learning approaches lack explainability. It introduces an end-to-end framework that first reduces the action space via a knowledge-driven predictor and SHAP-based parameter selection, and then employs a differentiable tree actor within an RL loop to generate configurations, followed by an interpretation tree to reveal the rationale. The approach achieves competitive performance on MySQL and Gbase8s, with slight improvements over existing RL-based tuners while offering transparent, actionable explanations for administrators. This combination of performance, stability, and explainability makes KnobTree a practical tool for real-world database tuning and a foundation for further research in explainable intelligent configurations.

Abstract

Databases are fundamental to contemporary information systems, yet traditional rule-based configuration methods struggle to manage the complexity of real-world applications with hundreds of tunable parameters. Deep reinforcement learning (DRL), which combines perception and decision-making, presents a potential solution for intelligent database configuration tuning. However, due to black-box property of RL-based method, the generated database tuning strategies still face the urgent problem of lack explainability. Besides, the redundant parameters in large scale database always make the strategy learning become unstable. This paper proposes KnobTree, an interpertable framework designed for the optimization of database parameter configuration. In this framework, an interpertable database tuning algorithm based on RL-based differentatial tree is proposed, which building a transparent tree-based model to generate explainable database tuning strategies. To address the problem of large-scale parameters, We also introduce a explainable method for parameter importance assessment, by utilizing Shapley Values to identify parameters that have significant impacts on database performance. Experiments conducted on MySQL and Gbase8s databases have verified exceptional transparency and interpretability of the KnobTree model. The good property makes generated strategies can offer practical guidance to algorithm designers and database administrators. Moreover, our approach also slightly outperforms the existing RL-based tuning algorithms in aspects such as throughput, latency, and processing time.

Figures (11)

• Figure 1: The overall model of KnobTree. The parameter selection module requires only a small amount of data to construct a predictive model, filtering parameters based on Shapley values. The tuning module consists of a tree-shaped actor network and a standard critic network. The explanation module transforms the tuning module's actor network into an explanation tree.
• Figure 2: The workflow of parameter selection. (a) Gathering expert knowledge; (b) Transforming this knowledge into a prediction tree; (c) Constructing a dataset and training the prediction model; (d) Analyzing parameter importance using Shapley values.
• Figure 3: Structure of expert experience tree. Internal nodes (blue) correspond to rules, while leaf nodes (green) correspond to predicted performance.
• Figure 4: Parameter screening using SHAP values on MySQL. The red box highlights the nine most important parameters selected through the analysis, representing key influences on database performance.
• Figure 5: Explainable tuning model structure. The overall architecture aligns with other RL-based tuning methods, but the actor is implemented by a differentiable decision tree.