Today's networks lack the support to satisfy the highly diverse and fast changing demands of emerging applications and services. The paradigms Network Virtualization (NV) and Software-Defined Networking (SDN) can potentially overcome this impasse. The virtualization of software-defined networks is expected to bring dynamic resource sharing with guaranteed performance through NV and programmability through SDN; for the first time, tenants can program their requested network resources according to their service demands in a timely manner. However, the virtualization of SDN-based networks introduces new challenges for operators, e.g., a virtualization layer that provides low and guaranteed control plane latencies for tenants. Moreover, tenants' expectations range from a fast, nearly-instantaneous provisioning of virtual networks to predictable operations of virtual networks. With this paper, we give a comprehensive overview of the thesis, which can be split into three parts - a journey in three acts. The thesis first presents a measurement procedure and a flexible virtualization layer design for the virtualization of software-defined networks. Focusing on the control plane, it introduces mathematical models for analyzing four virtualization layer architectures. Third, for a fast and efficient virtual network provisioning on the data plane, the thesis proposes optimization systems using Machine Learning and Neural Computation.     «

Today's networks lack the support to satisfy the highly diverse and fast changing demands of emerging applications and services. The paradigms Network Virtualization (NV) and Software-Defined Networking (SDN) can potentially overcome this impasse. The virtualization of software-defined networks is expected to bring dynamic resource sharing with guaranteed performance through NV and programmability through SDN; for the first time, tenants can program their requested network resources according t...     »