Authors: Dhavale Priyal Sandesh, Dhavale Pranjal Sandesh, Gaikar Aditi Bhausaheb, Dr. V. M. Dhede, Prof. M. S. Nalawade
Abstract: The rapid growth of online services, digital education and large-scale artificial intelligence has increased the electricity demand of cloud data centres, where servers, storage, networks and cooling equipment run around the clock. Rule-based management policies, which react to fixed thresholds, often leave servers underused and cooling plants over-provisioned. Artificial intelligence (AI) offers a different approach: models learn from operational data, anticipate demand and choose actions that save energy while protecting service quality. This review examines how AI is applied to four tasks in a green data centre: forecasting workload, allocating and consolidating virtual machines, scheduling tasks with awareness of energy and carbon, and controlling cooling systems. Statistical forecasting, recurrent neural networks, reinforcement learning and data-driven cooling control are explained and compared through representative studies. An AI-based green-cloud framework is then described that combines a learning-based demand predictor, an energy- and carbon-aware scheduler, a cooling optimiser and a renewable-supply monitor in a closed control loop. A simulation-based evaluation plan using energy, response time, utilization and carbon emission as metrics is outlined. Challenges including data quality, model interpretability, safety of autonomous control and the energy cost of the AI models themselves are discussed.
DOI: https://doi.org/10.5281/zenodo.23879779
