End-to-End Network Slicing for 5G&B Wireless Software-Defined Systems

As a key enabling technology for 5G&B systems, network virtualization allows multiple service providers to simultaneously and independently serve their users via virtualized network slices. However, this innovative technology cannot slice wireless resources without a paradigm shift in existing hardware-based architectures. In this paper, end-to-end network slicing is treated from a perspective of wireless software-defined networking architectures. It jointly optimizes all communication functionalities in both radio access and core networks to ensure optimal data throughput and congestion-free systems. First, based on software- defined cellular architectures, the idea of end- to-end (across access and core network domains) virtualization is introduced with dedicated control units, including high-level controllers and local baseband servers. Next, a stochastic utility-optimal virtualization problem is formulated, which jointly optimizes congestion control, flow routing, and power slicing to maximize the total incoming rates of wireless/wired flows, while satisfying the flow- queue stability and system-level constraints. After transforming the virtualization problem into a tractable form, an iterative network slicing algorithm is proposed that employs a primal-dual Newton method with quadratic convergence and achieves resource-efficient virtualization via control-unit coordination. Numerical results validate the efficacy of our solution, facilitating the 5G&B infrastructure-as-a-service.