Energy-aware resource management in virtual Radio Access Networks (vRANs)

The evolution beyond 5G (B5G) and towards 6G is presently starting off, and it is an opportunity not to be missed to introduce a neat turning point for cleaner and greener mobile communication systems. Any credible sustainability solution for 6G radio access must be designed in full compliance with the specificities of the ever-evolving Radio Access Network (RAN) architecture, which is embracing the virtualized RAN (vRAN) concept, combining, 
disaggregation, virtualization, cloudification and support for network slicing at the very edge of the network. By running RAN functionalities as Virtual Network Functions (VNFs) over commodity hardware on a (edge) cloud infrastructure, vRANs provide clear gains in resource pooling and multiplexing, as well as from coordinated processing.


In this context, the first objective of the proposed PhD thesis is to build open datasets of the power consumption at the User Equipment (UE) as well as in the energy-hungry vRAN infrastructures using the OpenAirInterface (OAI)-based 5G experimental testbed available at ÉTS-Montréal. These datasets will be then leveraged using machine learning (ML) to characterize the energy requirements of different VNFs configuration scenarios. We will then propose ""green network slicing"" paradigm, allowing to create logical networks (i.e. network slices), while ensuring the sustainability of the mobile network. In particular, we will propose ML models for green network slice capacity prediction at the 5G-RAN, and algorithms for green network slice admission control for energy-aware resource management in vRANs.