Energy-Aware Task Scheduling and Load Balancing in Cloud Computing Using AI
Shriya Sahu, Prerna Verma
Abstract
Energy-efficient task scheduling and load balancing in dynamic cloud computing environments remain critical challenges due to unpredictable workloads and high energy consumption. This study proposes a novel hybrid AI-driven framework that synergistically integrates Multi-Agent Collaborative Reinforcement Learning (MACRL), the bio-inspired Salp Swarm Algorithm (SSA), Neuroevolution of Augmenting Topologies (NEAT), Energy-Aware Task Scheduling with Deadline (EATSD), Merging Multiple Fused Learning Data Elements (MMFLDE), and Zeroth-Order Optimization (ZOO) to optimize resource allocation and energy consumption. The EATSD mechanism combined with NEAT enables adaptive and energy-efficient task scheduling by evolving neural network architectures. Simulation results conducted in MATLAB demonstrate that the proposed method achieves a superior efficiency score of 95% and an energy efficiency value of 2.26 bits per joule, surpassing existing approaches in energy savings, task completion times, and resource utilization. The novelty of this work lies in the synergistic integration of multiple advanced AI techniques into a unified framework addressing energy-aware task Unlike prior approaches focusing on individual algorithms, this hybrid method leverages complementary strengths to optimize resource allocation, adapt workload fluctuations, and minimize energy consumption, delivering enhanced performance and scalability. This comprehensive framework addresses limitations of individual methods and offers a significant advancement toward sustainable and intelligent cloud infrastructure management.