Doctor of Philosophy
Electrical and Computer Engineering
Dr. Abdallah Shami
The advent of next-generation networks has ushered in an era of enhanced performance requirements for our networks. As such, Network Service Providers are tasked with improving their networks to adhere to these performance requirements while aligning their operations with internal objectives. To this end, the Management and Orchestration of next-generation networks becomes an increasingly important topic to accomplish both of these goals. The work presented in this thesis targets several areas of modern networks, including Virtualised Networks, Multi-Access Edge Computing Networks, Optical Transport Networks, and 5G Core Networks. The methods employed in this thesis span the fields of operations research and intelligence and include optimization problem formulation (deterministic and robust), machine learning techniques (supervised, unsupervised, deep, federated), dimensionality reduction (PCA), and low-complexity heuristic solutions. These methods are used to address problems such as Vehicular and Next-Generation Service Placement, Virtualised Network Function Placement, Multi-layer Traffic Grooming and Infrastructure Placement, Core Network Traffic Analysis and Characterization, Core Network Intelligence Integration, Slice-Level Performance Metric Monitoring and Forecasting, and Model Drift Detection and Adaptation. By addressing the aforementioned problems from all aspects of the network, the end-to-end Quality of Service delivered to the end user can be improved. At the same time, the operational costs of the Network Operator can be reduced through proactive measures used to reduce the Operational Expenditures, specifically reactive maintenance. By focusing on aspects such as robustness, reliability, and resilience, the overall health of the network increases and its ability to survive adverse and unexpected conditions while maintaining consistent performance increases. The work presented in this thesis aims to help Network Operators revolutionize their practices to adapt to the changing networking landscape and position themselves in such a way that they can capitalize on the opportunities and potential of next-generation networks, services, systems, and applications.
Summary for Lay Audience
The introduction of 5G networks has revolutionized networking technologies worldwide. In order to keep up with the rapid pace of development, Network Service Providers must think of new ways to manage their networks. One of these ways is to introduce intelligence through tools such as Machine Learning. However, despite our best efforts, unforeseen changes stemming from uncertainty in a highly dynamic system can affect the quality of the network management solutions. To combat this, methods such as robust optimization, which actively considers uncertainty and protects the solution in adverse network conditions, can be leveraged. Furthermore, unforeseen events and changes in user behaviour can negatively affect the network and, specifically, the machine learning implementations. To address this, the development of frameworks that monitor and assess the performance of these implementations and signal when a significant change occurs is required. Signalling when a change has occurred is only half the job, and methods to adapt to these changes and maintain system performance are required. The work presented in this thesis addresses network management across various network segments, including large-scale core networks, resource-constrained edge networks, and high-speed optical networks. By understanding the characteristics of the core network, we can optimize the placement of services, which ultimately improves performance. By considering multiple ways to deliver a service, we can improve the reliability of the service and prevent outages. By aggregating network traffic, we can improve the utilization of our infrastructure and eliminate unnecessary costs. By developing methods to reduce the complexity of a solution, we can reduce computational costs and the time required to achieve said solution. Finally, by exploring different ways to improve service quality and reduce operation costs, we can help transition into the new era of networking with new and exciting use cases, applications, and opportunities. The work presented in this thesis addresses each of the above-mentioned points in an effort to help adapt current networks and networking practices to the requirements of future networks and unlock the true potential of next-generation networks and systems.
Manias, Dimitrios Michael, "Robust and Intelligent Management and Orchestration of Next-Generation Networks and Systems" (2023). Electronic Thesis and Dissertation Repository. 9361.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Available for download on Friday, June 27, 2025