Thermally Aware Machine Learning for Unified and Energy-Efficient Resource Management in Cloud Data Centers

This study presents a thermally aware machine learning approach aimed at enhancing energy efficiency and resource management within cloud data centers. By analyzing multiple system features such as CPU usage, memory consumption, network bandwidth, disk I/O, energy consumption, and service latency the proposed model predicts workload intensity categorized into low, medium, and high levels. Task priorities and optimized resource allocations, represented by the number of virtual machines or containers, are integrated into the framework to ensure balanced performance and energy conservation. Utilizing ensemble learning algorithms like Random Forest and XGBoost, the approach effectively models complex relationships between resource usage and energy consumption. The implementation employs a Python backend with the Flask framework, supported by a responsive front-end developed in HTML, CSS, and JavaScript. This solution facilitates adaptive and efficient allocation of cloud resources while minimizing power consumption, thus addressing the critical challenge of sustainability in large-scale data centers. The methodology demonstrates robust performance in predicting workload demands and optimizing resource distribution, which collectively contribute to reduced energy use and improved service latency. This integration of machine learning with resource management represents a significant step toward greener cloud computing infrastructures.

Keywords:
Energy-efficient resource management, cloud data centers, machine learning, workload prediction, Random Forest, XGBoost, resource allocation optimization, CPU utilization, energy consumption, service latency, Flask, cloud sustainability.