Title: SDN-Enabled Management of Access Infrastructures and Cloudbased Platforms
Supervisors
Principal supervisor: Associate Professor Michael Berger
Co supervisor: Associate Professor José Soler
Evaluation Board
Associate Professor Henrik Wessing, Department of Fotonics, DTU
Professor Andreas J. Kassler, Karsltad University, Sweden
Associate Professor Qi Zhang, Aarhus University, Denmark
Master of the Ceremony
Professor Søren Forchhammer, DTU Fotonik
Abstract
5th Generation (5G) networks aim to increase network convergence by allowing a single network architecture to serve any traffic type, while simultaneously, meeting their respective requirements. As a result, 5G networks must be characterized by flexibility, dynamicity, and determinism. Flexibility means that the network should scale its resources depending on the current load to ensure performance without resource overprovisioning. Dynamicity is the ability to modify the behavior of the network following a programmatic approach instead of relying on static (preconfigured) configurations. Finally, if 5G is to cater for mission-critical/time-sensitive applications such as industrial automation, it should provide deterministic and timely treatment for packets of these traffic types. The reason a new network architecture is required is that the current networking paradigm is based on static configurations and monolithic service deployments and thus lacks the characteristics mentioned above.
Recent years have seen the emergence of three networking paradigms that, if used appropriately, can provide high degrees of flexibility, dynamicity, and determinism. Namely, these are the Software Defined Networking (SDN), the Network Function Virtualization (NFV), and the Time Sensitive Networking (TSN) paradigms. By separating the network control plane from the data plane and logically centralizing it into an SDN Controller (SDNC), SDN allows for the dynamic/programmatic control of the network infrastructure. NFV, on the other hand, provides flexibility and scalability concerning service placement through virtualization and orchestration mechanisms, as network services can be instantiated, configured, and scaled on-demand. Finally. TSN guarantees the deterministic and timely treatment of time-sensitive traffic, without the need for a separate network infrastructure. The contributions of this Ph.D. thesis are towards these three paradigms and specifically towards investigating their applicability in different scenarios and identifying and addressing some of their limitations.
To provide deterministic treatment of time-sensitive traffic, TSN utilizes a scheduling-based approach in which a network-wide schedule controls the transmission capabilities of each egress port in the network. Chapter 3, of this thesis, provides two contributions to this scheduling problem; 1) It proposes a heuristic methodology for generating the network-wide schedules and 2) it investigates how the forwarding decisions for time-sensitive traffic can affect the quality of these schedules. As part of future work, this chapter also presents a network architecture which integrates TSN into an SDN/NFV framework.
As the default location for placement of virtualized service components, Data Centers (DC) will be an integral part of any 5G deployment. Given the stringent characteristics of 5G services, it is vital to identify and address any limitations of the SDN paradigm concerning a DC environment. Chapter 4 of this thesis identifies three of these limitations and presents solutions that can mitigate them. Specifically, chapter 4 provides the following contributions; 1) it proposes and presents two SDN-based network policy framework solutions that facilitate management of heterogeneous network infrastructures, 2) it investigates the applicability of Traffic Engineering (TE) in the context of SDN, and 3) it identifies the issue of limited capacity in network devices concerning the storage of network flow rules and proposes two solutions.
Finally, 5G deployments are expected to comprise a variety of Management and Operation (MANO) platforms as the different 5G services will require different support functions and guarantees. This multi-platform characteristic can increase the overall complexity of managing a 5G network, as each platform can bring its unique Application Programming Interfaces (API), protocols, and workflows. Chapter 5 of this thesis considers this issue and presents and validates a centralized and technology-agnostic framework for the deployment and management of 5G platforms and services. Moreover, chapter 5 investigates two additional research items. The first identifies and addresses limitations in the deployment process of a current state-of-the-art 5G platform solution, and the second identifies a set of requirements that SDNCs should meet to be considered as ready for the next-generation of 5G platforms.