This thesis addresses the deployment of energy- and cost-efficient yet reliable virtualized Radio Access Networks (vRANs) by overcoming the inefficiencies of the current industrial solution that relies on energy-hungry and expensive Hardware Accelerators (HAs) dedicated to individual Distributed Unit (DU) for processing. The proposed solutions include ECORAN, a multi-agent contextual bandit ML algorithm operating in the Near-Real-Time RAN Intelligent Controller (Near-RT-RIC) to offload DU workloads to either GPUs or CPUs, saving energy while preserving reliability and improving c ost-efficiency through DU centralization; YinYangRAN, a system in the Non-Real-Time-RIC that supervises GPU resource multiplexing to ensure reliable DU processing while maximizing throughput for concurrent ML services; and CloudRIC, a real-time brokering system that coordinates centralized access for multiple DUs to a heterogeneous pool of HAs and CPUs, ensuring vRAN reliability and cost and energy efficiency while adapting to milliseconds scale traffic dynamics. Experimental evaluations show that these solutions achieve up to 40% energy savings and 60x cost gains with ECORAN, improve vRAN reliability by over 50% with YinYangRAN, and deliver 3x and 15x average gains in energy and cost efficiency with CloudRIC, compared to the industry-standard solution, while maintaining 99.999% reliability.

About Leo Lo Schiavo

Leonardo Lo Schiavo is a Ph.D. student of the Networks Data Science Group at IMDEA Networks Institute since May 2020. He received his Bachelor’s Degree in Computer Science Engineering from the University of Catania in 2016 and a double Master’s Degree in Communications and Computer Networks Engineering from Politecnico di Torino and Politecnico di Milano in 2018, holding the research for his Master’s Thesis in the Multi-Access Edge Computing (MEC) team at Telecom Italia in Turin, where he focused on developing MEC solutions for safety applications in vehicular networks. His current main research interests include Network Functions Virtualization (NFV), RAN virtualization, Deep/Reinforcement Learning.

PhD Thesis Advisor:  Marco Fiore & Andres Saavedra Garcia

University: Universidad Carlos III de Madrid

Doctoral Program: Telematic Engineering

PhD Committee members:

• President: Claudio Fiandrino, Research Assistant Professor at IMDEA Networks Institute
• Secretary: Carla Fabiana Chiasserini, Full Professor at Politecnico di Torino and Chalmers University of Technology
• Panel member: Eduardo Baena, Postdoctoral Fellow at Northeastern University