Advanced Join Query Optimization Using Firefly and Reinforcement Learning Techniques on TPC-H Dataset

Abstract

Join query optimization is a critical component of database management systems (DBMS), significantly influencing their performance and efficiency. This study delves into advanced optimization techniques by employing the Firefly Algorithm and its hybrid integrations with Deep Q-Network (DQN) and Double Deep Q-Network (DDQN) methodologies. Utilizing the TPC-H benchmark dataset, we rigorously evaluate the efficacy of these algorithms in optimizing complex join queries. The Firefly Algorithm, inspired by the luminescent communication of fireflies, serves as a powerful metaheuristic optimization technique, adept at navigating vast search spaces. To augment this method, we incorporate reinforcement learning via DQN and DDQN, enhancing the algorithm's capability to balance exploration and exploitation during the optimization process. Our empirical analysis reveals substantial performance gains with the hybrid DQN-Firefly and DDQN-Firefly approaches compared to the standalone Firefly Algorithm. These findings underscore the potential of these hybrid methods for practical implementation in database management systems, promising improved query optimization and overall system performance.