Real Time Trading Of Mobile Resources In Beyond 5G Systems

5G is expected to offer download speeds as high as 1 GBps and latency lower than 10 ms that lead to the need for enormous investments. However, many small operators simply do not possess the necessary revenue in order to deploy the required infrastructure, while the rich operators are unwilling to burden this extreme cost due to the very long return of investment duration. Consequently, the technoeconomic pressure is forcing mobile operators to make pivotal changes in their modus operandi. A simple solution is to extent the conventional infrastructure sharing strategy to cover the active network components. However, despite the offered cost efficiency, these sharing approaches rely on well-defined service level agreements (SLA) that cover long-time intervals (e.g. years). However, this static model cannot provide the envisioned flexibility and efficiency for 5G. On the other hand, while the aforesaid techno-economic pressure is forcing the operators to share their networks, meanwhile the service heterogeneity demands revolutionary changes in the network management. The traditional solution of optimizing the complete network for a particular service type is no longer applicable due to the conflicting requirements posed by different services. A way out is to vertically group network resources, aka. slicing the network, in order to create virtual dedicated networks per service. This way, each resource group (i.e. slice) can be customized to serve a service in the best possible way. The conventional network provisioning technique, i.e. slicing the network based on statistical information, has a tendency towards over-provisioning the network. On the other hand, dynamic network slicing can increase efficiency but inter-tenant relatedness in a dynamic negotiation and resource allocation framework is still open issue. Consequently, the main research question in this PhD thesis revolves around how to achieve flexibility and efficiency in a shared mobile network. In particular, this thesis targets answering, 1) how the network resources can be dynamically and flexibly shared, 2) how the tenants can differentiate their services in a shared infrastructure, 3) what the long- and short-term implications of anticipatory network sharing and resource trading are. The proposed dynamic negotiation and resource allocation framework proposes a novel SLA understanding that can allow the operators to renegotiate in very short time scales (i.e. seconds). Moreover, we demonstrate how to exploit anticipatory information regarding the user's achievable rates. Lastly, we present a novel self-dimensioning algorithm that exploits short-term observations on the traffic demand to scale the network capacity. A number of simulations has been performed in order to investigate the efficiency of the proposed framework.