Energy-efficient SDN routing with deep learning for data center networks

The exponential growth of cloud storage services has led to a sharp increase in energy consumption across data center networks. To address this challenge, we propose a Software-Defined Networking (SDN)-based architecture focused on energy-efficient routing. The core of the system is a novel routing algorithm that combines Shortest Path First (SPF) principles with a deep learning decision module to dynamically select low-energy paths while maintaining quality of service. Network devices exchange status information using a custom extension of the OpenFlow protocol, allowing the SDN controller to make data-driven routing decisions based on real-time energy and performance metrics. The proposed architecture is validated through simulations and experiments on real network topologies, demonstrating energy savings under both synthetic and production-like traffic conditions. Experimental validation using both simulated and real network topologies shows that the proposed GMPC approach reduces energy consumption compared to standard OSPF routing, achieving energy savings of up to 7% in simulated mesh topologies and approximately 5% in real network environments, while maintaining QoS within acceptable thresholds. • Hybrid SDN routing combining MPC and OSPF with deep learning selection. • Neural model selects low-energy paths using real-time ESM and traffic data. • GMPC architecture reduces energy consumption without degrading QoS. • Real-network tests validate consistent energy reductions under heavy load.