Home / Current Issue / Paper 1719075
Cooperative Multi-Agent Negotiation for Closed-Loop Database Tuning and Index Optimization
Subject area: Science,Engineering and Technology · Area of research: Cloud Database
DOI: https://doi.org/10.64388/IREV9I12-1719075
Abstract
Modern database systems operate under increasingly dynamic workloads that render static configuration and manual tuning obsolete. Closed-loop database tuning, where the system automatically observes the performance metrics, diagnoses the bottlenecks and applies corrective actions, has emerged as a promising paradigm for self-managing databases. But the tuning problem is inherently multi-objective and multi-component: knob configuration, index selection, query rewriting and materialized view maintenance each need specialized optimization strategies which can conflict with each other. In this paper, we propose a cooperative multi-agent negotiation framework in which autonomous agents are responsible for a specific tuning dimension and engage in structured negotiation protocols to reach collectively optimal decisions. We formalize the negotiation as a constrained multi-objective optimization problem, design a coalition formation mechanism based on cooperative game theory, and evaluate the framework on the TPC-H and TPC-DS benchmarks against state-of-the-art single-agent and heuristic baselines. Experimental results show that the proposed framework improves the aggregate throughput by 18-34% and the tail latency by 22-41% over existing approaches while being stable against the workload phase transitions. The paper presents a principled theoretical basis for multi-agent database tuning, and a practical architecture that can be deployed in production environments.
Keywords
Multi-Agent Systems, Database Tuning, Index Optimization, Closed Loop Control, Cooperative Game Theory, Self Managing Databases, Query Optimization
How to cite this paper
@article{1719075,
author = {Shahid Moosa},
title = {Cooperative Multi-Agent Negotiation for Closed-Loop Database Tuning and Index Optimization},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {12},
pages = {2120-2127},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1719075.pdf},
abstract = {Modern database systems operate under increasingly dynamic workloads that render static configuration and manual tuning obsolete. Closed-loop database tuning, where the system automatically observes the performance metrics, diagnoses the bottlenecks and applies corrective actions, has emerged as a promising paradigm for self-managing databases. But the tuning problem is inherently multi-objective and multi-component: knob configuration, index selection, query rewriting and materialized view maintenance each need specialized optimization strategies which can conflict with each other. In this paper, we propose a cooperative multi-agent negotiation framework in which autonomous agents are responsible for a specific tuning dimension and engage in structured negotiation protocols to reach collectively optimal decisions. We formalize the negotiation as a constrained multi-objective optimization problem, design a coalition formation mechanism based on cooperative game theory, and evaluate the framework on the TPC-H and TPC-DS benchmarks against state-of-the-art single-agent and heuristic baselines. Experimental results show that the proposed framework improves the aggregate throughput by 18-34% and the tail latency by 22-41% over existing approaches while being stable against the workload phase transitions. The paper presents a principled theoretical basis for multi-agent database tuning, and a practical architecture that can be deployed in production environments.},
keywords = {Multi-Agent Systems, Database Tuning, Index Optimization, Closed Loop Control, Cooperative Game Theory, Self Managing Databases, Query Optimization},
month = {June},
doi = {https://doi.org/10.64388/IREV9I12-1719075}
}