Enhancing Performance, Security, and Management in Network Function Virtualization

In an era of ubiquitous connectivity, various new applications, network protocols, and online services (e.g., cloud services, distributed machine learning, cryptocurrency) have been constantly creating, underpinning many of our daily activities. Emerging demands for networks have led to growing traffic volume and complexity of modern networks, which heavily rely on a wide spectrum of specialized network functions (e.g., Firewall, Load Balancer) for diverse purposes. Although these (virtual) network functions (VNFs) are widely deployed, they are instantiated in an uncoordinated manner failing to meet growing demands of evolving networks. In this dissertation, we argue that networks equipped with VNFs can be designed in a fashion similar to how computer software is programmed today. By following the blueprint of modularization, networks can be made more efficient, secure, and manageable.