Date of Award

Fall 10-1-2021

Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical Engineering (ENAS)

First Advisor

Tassiulas, Leandros


With new possibilities brought by the Internet of Things (IoT) and edge computing, the traffic demand of wireless networks increases dramatically. A more sophisticated network management framework is required to handle the flow routing and resource allocation for different users and services. By separating the network control and data planes, Software-defined Networking (SDN) brings flexible and programmable network control, which is considered as an appropriate solution in this scenario.Although SDN has been applied in traditional networks such as data centers with great successes, several unique challenges exist in the wireless environment. Compared with wired networks, wireless links have limited capacity. The high mobility of IoT and edge devices also leads to network topology changes and unstable link qualities. Such factors restrain the scalability and robustness of an SDN control plane. In addition, the coexistence of heterogeneous wireless and IoT protocols with distinct representations of network resources making it difficult to process traffic with state-of-the-art SDN standards such as OpenFlow. In this dissertation, we design a novel architecture for the wireless network management. We propose multiple techniques to better adopt SDN to relevant scenarios. First, while maintaining the centralized control plane logically, we deploy multiple SDN controller instances to ensure their scalability and robustness. We propose algorithms to determine the controllers' locations and synchronization rates that minimize the communication costs. Then, we consider handling heterogeneous protocols in Radio Access Networks (RANs). We design a network slicing orchestrator enabling allocating resources across different RANs controlled by SDN, including LTE and Wi-Fi. Finally, we combine the centralized controller with local intelligence, including deploying another SDN control plane in edge devices locally, and offloading network functions to a programmable data plane. In all these approaches, we evaluate our solutions with both large-scale emulations and prototypes implemented in real devices, demonstrating the improvements in multiple performance metrics compared with state-of-the-art methods.