Optimizing edge computing resources towards greener networks and services

The world is at the dawn of a new era, characterized by a large number of connected devices and a resulting massive availability of networked data. In this scenario, multi-access edge computing (MEC) is the candidate reference paradigm to provide mobile users with low latency processing and storage services. In contrast with mobile cloud computing (MCC), MEC entails the deployment of computing facilities closer to the end devices, thus becoming a key enabler for applications such as augmented reality, tactile Internet, smart home, healthcare monitoring, connected cars, online gaming, etc. However, the problem of the energy efficiency of the decentralized MEC infrastructure arises. In this doctoral thesis, the management of the MEC platform is optimized to reduce the global carbon footprint of the network, i.e., to maximize the use of renewable energy resources (RERs) for the initial placement and the subsequent execution and offloading of jobs. The main body of the thesis is organized into three chapters: the first one tackles the green energy management of edge servers equipped with a battery in a hierarchical MEC network and the electricity trade with the power grid; the second chapter considers a vehicular scenario where vehicles' trajectories are proactively tracked when migrating the users' computing tasks towards increasing the energetic efficiency of this process; the third chapter presents a comparison of the two decentralized optimization approaches designed, based on message passing. The results show that the proposed optimization frameworks, based on model predictive control (MPC), can significantly reduce the carbon footprint of the edge network when compared to simple heuristics and other approaches in the scientific literature. The designed algorithms can reach almost complete carbon neutrality in a vast range of network conditions.