Optimized Cloud Load Balancing Using Deep Q-Learning With Inverted S-Shaped Hierarchical Reverse Superb Fairy-Wren Optimization

Abstract

Cloud computing offers a scalable and cost-effective platform by providing on-demand access to shared computational resources. However, effective load balancing is essential to maintain optimal performance and maximize resource utilization for ensuring even distribution of network traffic across servers, preventing overload, improving response time and enhancing the system reliability. This manuscript proposes the Deep Q-Network with Inverted S-shaped Hierarchical Reverse Superb Fairy-wren Optimization (DQN with ISHR-SFO) model for adaptive load balancing on cloud computing. DQN utilizes Reinforcement Learning (RL) to predict optimal task-to-VM allocations based on states and rewards. The traditional SFO algorithm improved global exploration and local exploitation by incorporating inverted S-shaped escape energy and hierarchical reverse learning to prevent premature convergence and improve the search diversity. The proposed model obtained lower execution time of 1443s, makespan of 8495s, energy consumption of 35.25 J, high throughput of 120 kbps and 99.12% resource utilization. Experimental evaluation using CloudSim determines that the proposed model effectively minimizes energy consumption, makespan and execution time, while increasing throughput and resource utilization compared to traditional algorithms.