So far there still lacks an effective methodology to leverage mmWave to accomplishflexible and efficient ultra-dense cellular networking for 5G networks. To this end, thorough research will be carried out in this project to investigate networking mechanisms, algorithms, and protocols for mmWave-based ultra-dense cellular networks, and ultimately develop a complete mmWave-based networking framework for ultra-dense cellular networks. By considering the characteristics of mmWave communications and ultra-dense cellular networks, this project starts with a unique design of heterogeneous networking architecture, which is distinct with features such as hybrid mesh networking for integrating backhaul and access networks, control and management of mmWave networks via macro cellular networks, and separation between user plane and control plane. Based on this architecture, a complete methodology for mmWave-based ultra-dense cellular networking is established as follows. For backhaul networks, key physical and medium access control (MAC) mechanisms are developed first. These mechanisms include beam switching and configuration for dynamic and static beams, fast detection of link degradation, and quick response to link failure. Besides, a novel resource management algorithm is derived to consolidate centralized optimization and distributed scheduling, with its long-term performance proved to be optimal. Based on these mechanisms and the resource management algorithm, networking protocols are designed to achieve flexible mesh networking for mmWave backhaul networks. For access networks, key mechanisms are developed first to handle mmWave mobile communications. Such mechanisms include fast cell discovery, quick beam alignment, and efficient mobility management. They work cooperatively with virtual cells that are created from multiple small cells and maintained dynamically for mobile users. Based on these mechanisms and virtual cells, algorithms and protocols are designed to deliver stable services to users moving across mmWave access networks. Finally, a simulation platform will be developed, through which the research results of this project are validated. The effectiveness, advantages, and performance gains of the entire methodology developed in this project will also be demonstrated by running typical networking scenarios on the simulation platform.