Abstract :
[en] The telecom industry foresaw a constant change over the decades. Alongside the standardization of 5G, not only the traffic demand increased, but also the different requirements of the provided services, e.g., latency, traffic load and pattern, datarate. To match this trend, the traditional telecom infrastructure has been revolutionized, going from the "one-size-fits-all" model to a shared approach, where the infrastructure is shared among enterprise customers with even very different requirements. In this context, a paradigm, named network slicing, has considerably attracted the network operators. Network slicing is a successful enabling model for the 5G and beyond networks because it allows operators to tailor their networks based on the end use case, release unused functionalities and resources and dynamically assign them to different customers. In addition, 6G networks are advertised to bring the terrestrial and satellite networks closer, to work in a coordinated way and provide ubiquitous, heterogenous and reliable services. In this context, this thesis investigates the optimization of network slicing in an integrated satellite-terrestrial network. Well-known enabling technologies for network slicing, such as Software-Defined Networking (SDN), are included as a proof-of-concept SDN-based testbed to demonstrate and support the proposed optimization algorithms. As network slicing is a resource allocation problem, where virtualized resources are accommodated on the substrate network, we investigate this optimization problem that is well-known in the literature as Virtual Network Embedding (VNE). Firstly, we study the application of VNE to an integrated Medium Earth Orbit (MEO)terrestrial network with the objective of minimizing the traffic migrations, considering the existence of inter-satellite links (ISLs) too. As satellite handovers are unavoidable, we showed as including the minimization of traffic handovers in the objective function, brings to a gain in terms of traffic migrations and packet loss up to 2.5-5% compared to the traditional approaches. Secondly, we investigated the benefit of flexibly accommodating traffic demands without fully assigning the required resources while keeping the user satisfaction probability (USP) under control. Thanks to the SDN-based testbed, the traffic is generated and the statistics are collected to real-time match the need of each user. This showed an increase in the acceptance
ratio up to 11% compared to the baselines.
Lastly, as the previous two main chapters investigated the point-to-point connectivity, we expanded the work to the embedding of full slices for a combined Geostationary (GEO)-Low Earth Orbit (LEO) satellites and terrestrial network. We provided a flexible framework, for 6G use-cases with real network requirements, which operates based on prioritization, minimizes the migrations of slices when congestions occur over the substrate network and proactively manages the satellite handovers for each slice.
Finally, we discuss conclusive remarks and future research directions.
