Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Doctor of Philosophy

Program

Electrical and Computer Engineering

Supervisor

Shami, Abdallah

Abstract

The current trend in user services places an ever-growing demand for higher data rates, near-real-time latencies, and near-perfect quality of service. To meet such demands, fundamental changes were made to the front and mid-haul and backbone networking segments servicing them. One of the main changes made was virtualizing the networking components to allow for faster deployment and reconfiguration when needed. However, adopting such technologies poses several challenges, such as improving the performance and efficiency of these systems by properly orchestrating the services to the ideal edge device. A second challenge is ensuring the backbone optical networking maximizes and maintains the throughput levels under more dynamically variant conditions. A third challenge is addressing the limitation of placement techniques in O-RAN. In this thesis, we propose using various optimization modeling and machine learning techniques in three segments of network systems towards lowering the need for human intervention targeting zero-touch networking. In particular, the first part of the thesis applies optimization modeling, heuristics, and segmentation to improve the locally driven orchestration techniques, which are used to place demands on edge devices throughput to ensure efficient and resilient placement decisions. The second part of the thesis proposes using reinforcement learning (RL) techniques on a nodal base to address the dynamic nature of demands within an optical networking paradigm. The RL techniques ensure blocking rates are kept to a minimum by tailoring the agents’ behavior based on each node's demand intake throughout the day. The third part of the thesis proposes using transfer learning augmented reinforcement learning to drive a network slicing-based solution in O-RAN to address the stringent and divergent demands of 5G applications. The main contributions of the thesis consist of three broad parts. The first is developing optimal and heuristic orchestration algorithms that improve demands’ performance and reliability in an edge computing environment. The second is using reinforcement learning to determine the appropriate spectral placement for demands within isolated optical paths, ensuring lower fragmentation and better throughput utilization. The third is developing a heuristic controlled transfer learning augmented reinforcement learning network slicing in an O-RAN environment. Hence, ensuring improved reliability while maintaining lower complexity than traditional placement techniques.

Summary for Lay Audience

As the use of smart and connected devices continues to grow, there is a need for more reliable and faster networking infrastructure. This thesis focuses on addressing the challenges faced by 5G systems and their underlying infrastructure in three domains: user-adjacent edge computing, Open Radio Access Networks (O-RANs), and back-end Optical networking. The main goal is to automate these networks to improve their reliability and speed. To achieve this, the thesis presents optimization algorithms, heuristic strategies, and machine-learning models to enhance the performance of network orchestration, slicing, and spectrum allocation. By implementing these contributions, network service providers can offer more reliable, secure, and fast services. The resulting networks can support more diverse applications with increasing demands such as Autonomous Vehicles, Augmented Reality, Industrial Automation, Emergency Services, and Smart Grids. This thesis paves the way for a new era of networking with automated organization and self-heal capabilities that target zero-touch networks, where no human interaction is needed to maintain or fix the networks in cases of failures.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS