Sdn architecture for creating simple and flexible cellular and mobile networks

Résumé

The network softwarization paradigm is playing a key role in the evolution of next-generation cellular networks. Besides satisfying the requirements for the ever increasing demand for higher data rates, Fifth Generation (5G) networks aim to support a wide range of use cases for different vertical industries. This introduces an additional set of challenging requirements for cellular networks such as the need for supporting low latency communications and providing simultaneous connectivity to a massive number of devices. The contradictory requirements of the different use cases makes unfeasible the traditional static network design for peak cases. Thus, to support the specific requirements of each use case in an efficient manner, the network needs to be flexible and adaptable. For this reason, concepts such as Network Slicing and Multi-Access Edge Computing (MEC) become relevant in the network architecture. They are enabled by the network softwarization paradigm, which embraces the adoption of Software-Defined Networking (SDN) and Network Function Virtualization (NFV) for providing the required flexibility. They rely on splitting control and data planes, centralizing control plane decisions and decoupling Network Functions (NFs) from the underlying hardware. This thesis contributes to the broad field of network softwarization by continuing the current trends in the evolution of cellular networks. We study how the SDN paradigm can be leveraged both to introduce flexibility in cellular networks and also to simplify their architecture. The contributions aim at advancing the state-of-the-art in the cellular network softwarization research field by providing flexibility, automated orchestration and simplified control-plane communications in dynamic scenarios with the introduction of SDN in the network architecture. Our proposals consider interoperability as a major requirement in order to facilitate the integration with existing deployments. In addition, the proposals are experimentally evaluated in real-world testbeds, analyzing the challenges that arise in their implementation and integration. The contributions can be categorized along three main axis: dynamic MEC, automated orchestration and NF connectivity. In dynamic MEC, we first propose an SDN-based mechanism to support session and service continuity in dynamic MEC, which allows the transparent relocation of user communications from the core to edge resources without disrupting ongoing communications. We experimentally evaluate the proposal in a testbed implementation, where we validate its interoperability with existing deployments. Then, we formally model the problem of the dynamic deployment of anchor points and the assignment of users to them in order to minimize the latency perceived and also reduce the overhead introduced in the network. We propose algorithms that are capable of solving this problem efficiently. In automated orchestration, we propose a new Key Performance Indicators (KPI), which we denominate as Closed Loop Orchestration Delay (CLOD), for measuring the agility of NFV Management and Orchestration (MANO) platforms. We evaluate CLOD through testbed experiments comparing the overhead introduced by fully functional MANO platforms with a baseline (a custom development with the minimal functionality). We also proposed an architecture for conducting experiments about the migration of MEC applications in scenarios with user mobility. The architecture, which integrates MANO and MEC platforms, is also validated through testbed experiments. Finally, related to NF connectivity, we propose a new SDN-based solution for interconnecting NFs in service-based architectures, which introduces the benefits of indirect communication without a performance degradation. This solution can be directly integrated in 5G networks as an alternative to the solution defined by the standard: the Service Communication Proxy (SCP), thus simplifying the architecture while maintaining the functionality.