Date of Award
Fall 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical Engineering (ENAS)
First Advisor
Tassiulas, Leandros
Abstract
A “next-generation” of networks is emerging as the requirements for and capabilities of modern networked systems grow increasing more complex compared to their predecessors. These next-generation networks include cloud data centers with hardware disaggregation, mega-constellations of low-Earth orbit (LEO) satellites, and alternative fuel vehicle transportation networks. To ensure efficient design and operation, they require more sophisticated resource allocation strategies. Better resource allocation strategies unlock the full potential of next-generation network infrastructure by using resources more efficiently. These resource allocation strategies that improve efficiency are particularly relevant today as governments, companies, and researchers have begun placing greater emphasis on environmental sustainability. Sustainability is a philosophy that seeks balance between the immediate needs of networking applications and the health of global ecosystems. While sustainable development is necessary to improve long-term climate outlooks, it can also require significant financial investment and cause disruptions to existing services. In this dissertation, we introduce several resource allocation strategies that can achieve good performance at the large scales and complexities of real-world next-generation networks. The applications we consider relate to two major themes in sustainable development: improving the energy efficiency of existing networks and building new network infrastructure to facilitate long-term energy efficiency. More specifically, we consider a diverse set of next-generation networking applications and solve them using efficient deep reinforcement learning and integer linear programming techniques. We introduce an energy-efficient task scheduler that reduces energy usage in disaggregated cloud data centers. We design a computational offloading strategy for rural and remote Internet of things (IoT) applications that utilizes the computational resources in orbiting satellites. We formulate the capacitated refueling station location problem with routing to minimize the cost of building new alternative refueling stations subject to capacity limits at refueling stations due to congestion. Finally, we develop a network slicing algorithm that enables data-intensive remote sensing applications in non-terrestrial edge networks.
Recommended Citation
Nordlund, Nicholas, "Resource Allocation Problems for Efficient and Sustainable Next-Generation Networks" (2023). Yale Graduate School of Arts and Sciences Dissertations. 1187.
https://elischolar.library.yale.edu/gsas_dissertations/1187
