Enhancing Resource Management and Energy Consumption Forecasting in Multi-Cloud Environments with AI-driven Approaches
Jay Barach, Junaid Hussain Muzamal · 2024
This paper provides an in-depth analysis of optimizing resource allocation in multi-cloud environments through the integration of advanced machine learning and optimization techniques. We implemented a comprehensive strategy that combines CNN for predicting resource demands, Ant Colony Optimization (ACO) for optimizing initial resource allocation, PPO for dynamic adjustments, and GBM for forecasting energy consumption. The CNN model exhibited strong performance in demand prediction, with detailed metrics demonstrating its effectiveness across various scenarios. ACO significantly improved resource distribution efficiency, as evidenced by a reduction in active cloud regions. The system’s adaptive capabilities were further enhanced by Proximal Policy Optimization (PPO), which dynamically optimized resource allocation in response to real-time demand fluctuations. Additionally, the GBM model provided highly accurate energy consumption predictions, closely aligning with actual usage data and underscoring the potential for energy-efficient management in multi-cloud environments.