Optimal control policies with qos and infrastructure slicing for millimeter wave cellular networks

Resumo

Dense deployment of millimeter wave (mmWave) cells is crucial to face the capacity challenge of future cellular systems (5G and beyond). To fully harness their potential, wireless backhauling and infrastructure virtualization are key technological enablers. Wireless backhauling using mmWave connections between mmWave cells is a new scenario in the cellular paradigm and its success depends on a profound revision of Time Division Duplexing (TDD) and design of new routing schemes. Firstly, because the current cellular paradigm uses static TDD schemes that would be rather inefficient in a mmWave network due to the increased traffic dynamics. Secondly, because uplink (UL) and downlink (DL) transmissions cannot happen at the same time due to interference. In mmWave, however, this UL/DL separation is unnecessary thanks to high transmission directivity. Thirdly, because mmWave cellular networks will rely on new complex mesh topologies that are not considered in current cellular standards, not even in relaying operation mode, which would severely degrade the achievable capacity region of mmWave networks. Moreover, a new relaying mode should also incorporate an efficient mechanism to differentiate data traffic when routing. Therefore, optimal routing in these complex networks must be addressed. In the first part of this dissertation we contribute to address the challenges above. We provide opportunistic network control policies to exploit the network capacity region with completely dynamic TDD operation under arbitrary topologies. The proposed policies involve link scheduling and routing problems and we present analytical performance guarantees such as throughput-optimallity and end-to-end delay minimization. In addition, we incorporate Quality-of-Service traffic differentiation with optimal network-level delay. The theoretical results hold with the markovian behavior of the mmWave channel in time. Sharing resources using virtualization techniques increases flexibility at reduced costs. While spectrum scarcity is the most compelling factor to increase performance in below-6 GHz cellular systems, site density becomes more critical in mmWave due to the blockage problem. Therefore, infrastructure sharing emerges naturally as a solution in the new mmWave cellular paradigm. Sharing cell sites between operators contributes to reducing capital and operational expenditures since only a single mmWave network deployment is needed. At the same time, the large bandwidth and the high spatial diversity of mmWave cells still boost operator performance even sharing the base stations. In the second part of this dissertation we contribute to the analysis of infrastructure sharing among operators. In our approach, an infrastructure provider offers its mmWave base stations to several mobile operators in exchange for a payment. We propose a distributed infrastructure auction mechanism in which operators compete for base stations resources based on their own loads. In this manner, each operator can obtain a network infrastructure slice dynamically and on demand to serve its users. We believe that our control policies provide solid foundations for the development of real protocols. Our virtualization framework provides pricing and allocation mechanisms for generating dynamic network slices in scenarios where load conditions change rapidly at base stations due to mobility and intermittent access.