Abstract

The Fifth Generation (5G) era is touted as the next generation of mobile networks that will unleash new services and network capabilities, opening up a whole new line of businesses recognized by a top-notch Quality of Service (QoS) and Quality of Experience (QoE) empowered by many recent advancements in network softwarization and providing an innovative on-demand service provisioning on a shared underlying network infrastructure. 5G networks will support the immerse explosion of the Internet of Things (IoT) incurring an expected growth of billions of connected IoT devices by 2020, providing a wide range of services spanning from low-cost sensor-based metering services to low-latency communication services touching health, education and automotive sectors among others. Mobile operators are striving to find a cost effective network solution that will enable them to continuously and automatically upgrade their networks based on their ever growing customers demands in the quest of fulfilling the new rising opportunities of offering novel services empowered by the many emerging IoT devices. Thus, departing from the shortfalls of legacy hardware (i.e., high cost, difficult management and update, etc.) and learning from the different advantages of virtualization technologies which enabled the sharing of computing resources in a cloud environment, mobile operators started to leverage the idea of network softwarization through several emerging technologies. Network Function Virtualization (NFV) promises an ultimate Capital Expenditures (CAPEX) reduction and high flexibility in resource provisioning and service delivery through replacing hardware equipment by software. Software Defined Network (SDN) offers network and mobile operators programmable traffic management and delivery. These technologies will enable the launch of Multi-Access Edge Computing (MEC) paradigm that promises to complete the 5G networks requirements in providing low-latency services by bringing the computing resources to the edge of the network, in close vicinity of the users, hence, assisting the limited capabilities of their IoT devices in delivering their needed services. By leveraging network softwarization, these technologies will initiate a tremendous re-design of current networks that will be transformed to self-managed, software-based networks exploiting multiple benefits ranging from flexibility, programmability, automation, elasticity among others.

This dissertation attempts to elaborate and address key challenges related to enabling the re-design of current networks to support a smooth integration of the NFV and MEC technologies. This thesis provides a profound understanding and novel contributions in resource and service provisioning and scheduling towards enabling efficient resource and network utilization of the underlying infrastructure by leveraging several optimization and game theoretic techniques. In particular, we first, investigate the interplay existing between network function mapping, traffic routing and Network Service (NS) scheduling in NFV-based networks and present a Column Generation (CG) decomposition method to solve the problem with considerable runtime improvement over mathematical-based formulations. Given the increasing interest in providing low-latency services and the correlation existing between this objective and the goal of network operators in maximizing their network admissibility through efficiently utilizing their network resources, we revisit the latter problem and tackle it under different assumptions and objectives. Given its complexity, we present a novel game theoretic approach that is able to provide a bounded solution of the problem. Further, we extend our work to the network edge where we promote network elasticity and alleviate virtualization technologies by addressing the problem of task offloading and scheduling along with the IoT application resource allocation problem. Given the complexity of the problem, we propose a Logic-Based Benders (LBBD) decomposition method to efficiently solve it to optimality.