Unit:
Department of PhysicsLibrary of the School of Science
Author:
Alevizaki Viktoria-Maria
Dissertation committee:
Άννα Τζανακάκη, Αναπληρώτρια Καθηγήτρια στο Τμήμα Φυσικής, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Γεώργιος Τόμπρας, Καθηγητής στο Τμήμα Φυσικής, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Δήμητρα Συμεωνίδου, Καθηγήτρια στο Πανεπιστήμιο του Bristol
Μάρκος Αναστασόπουλος, Αναπληρωτής Καθηγητής στο Τμήμα Φυσικής, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Διονύσιος Ρεϊσης, Καθηγητής στο Τμήμα Φυσικής, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Σπύρος Δενάζης, Καθηγητής στο Τμήμα Ηλεκτρολόγων Μηχανικών και Τεχνολογίας Υπολογιστών, Πανεπιστήμιο Πατρών
Αθανάσιος Κοράκης, Καθηγητής στο Τμήμα Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, Πανεπιστήμιο Θεσσαλίας
Original Title:
Distributed Resource Management in Converged Telecommunication Infrastructures
Translated title:
Distributed Resource Management in Converged Telecommunication Infrastructures
Summary:
The fifth generation (5G) of wireless and mobile communications is expected to have a far-reaching impact on society and businesses beyond the information and communications technology (ICT) sector. 5G is aligned with the 4th industrial evolution, blurring the lines between the physical, digital, and biological spheres. A common design is necessary to accommodate all service types based on energy and cost efficiency. To address this, this PhD thesis adopts the idea of a universal 5G platform that integrates a variety of networking technologies (wireless and wired), and aims to develop mathematical tools, algorithms and protocols for the energy and operational optimization of this infrastructure and the services it provides. This infrastructure interconnects computing, storage and network components that are placed at different locations, using the concepts of programmable hardware (hardware-HW) and network software (network softwarisation). In this way, it enables the provision of any service by flexibly and efficiently mixing and matching network, computing and storage resources.
The thesis targeted four distinct contributions. All proposed contributions are implemented and investigated experimentally in a 5G open-source lab testbed. The first contribution focused on optimal function and resource allocation adopting the innovative 5G RAN architecture, that splits flexibly the baseband processing function chain between Remote, Distributed and Central Units. This enables access to shared resources provided by micro or large-scale remote data centers, without requiring resource ownership. To support this architecture, networks adopt the Software Defined Networking (SDN) approach, where the control plane is decoupled from the data plane and the associated network devices and is centralized in a software-based controller. In this context, the goal of the proposed approach was to develop effective optimization techniques that identify the optimal functional split, along with the optimal location and size of the SDN controllers. The second contribution concentrated on solving the User Plane Function (UPF) selection problem in 5G core networks. According to the SDN paradigm 5G core control plane functions manage the network, while UPFs are responsible for handling users’ data. Depending on the 5G RAN deployment option and the nature of the service, UPF nodes can be placed closer to the network edge, directing traffic to the Multi-access Edge Computing (MEC) servers hence reducing latency, or be placed deeper into the network directing traffic to central cloud facilities. In this context, a framework that selects the optimal UPF nodes to handle user’s traffic minimizing total service delay has been proposed. The third contribution pertained to service provisioning in upcoming telecommunication systems. 6G systems require novel architectural Quality of Experience (QoE) models and resource allocation strategies that can differentiate between data streams originating from the same or multiple User Equipment (UEs), respond to changes in the underlying physical infrastructure, and scale with the number of connected devices. Currently, centralized management and network orchestration (MANO) platforms provide this functionality, but they suffer scalability issues. Therefore, future systems are anticipated to operate in a distributed manner, allowing applications to directly intervene in relevant control processes to ensure the required QoE. The proposed approach focused on developing a flow assignment model that supports applications running on UEs. The final contribution of this thesis focused on the deployment of a 5G infrastructure that supports multi-tenant network slicing on demand. Sharing of the underlying physical infrastructure was achieved through the development of suitable interfaces for integrating different network components and the creation of appropriate descriptors for virtual 5G network functions (VNFs). By collecting and combining multiple VNFs, an end-to-end 5G Network Service (NS) can be obtained. Using a MANO platform, these NSs can be combined to instantiate and manage a 5G network slice.
Main subject category:
Science
Keywords:
5G, Software Defined Networking, Network Function Virtualization, Orchestration, Slicing, Edge Computing, 5G Transport Network, Distributed Resource Management, 5G Core, User Plane Function, QoS
Number of references:
168
