Software-Driven and Virtualized Architectures for Scalable 5G Networks

Cellular networks have become an integral part of our society. Today, the number of global 4G/LTE subscribers is around 1.2 billion with a maximum daily addition of 2 million devices since 2017. Also, diversity of devices supporting LTE is going beyond cell phones and is reaching to domestic robots, sensors, and cars. In this dissertation, we argue that it is essential to rearchitect 4G/LTE cellular core networks–sitting between the Internet and the radio access network–to meet the scalability, performance, and flexibility requirements of 5G networks. Today, there is a growing consensus among operators and research community that software-defined networking (SDN), network function virtualization (NFV), and mobile edge computing (MEC) paradigms will be the key ingredients of the next-generation networks.

Motivated by these trends, we design and optimize three core network architectures, SoftMow, SoftBox, and SkyCore, for different network scales, objectives, and conditions. SoftMoW provides global control over large-scale (e.g., continent-wide) core networks with the ultimate goal of enabling new routing and mobility optimizations. SoftBox attempts to enhance state management in core networks to enable low-latency, signaling-efficient, and customized services for mobile devices. SkyCore is aimed at realizing a core network for drone-based LTE networks that are going to serve first responders in the future. Network slicing makes it possible to deploy these solutions in parallel. To better support mobility and provide verifiable security, these architectures can use an addressing scheme that separates network locations and identities with self-certifying, flat and non-aggregatable address components. To benefit the proposed architectures, we designed a high-speed and memory-efficient router, called Caesar, for this type of addressing scheme.