Virtualized Network Infrastructures: Performance Analysis, Design and Implementation

In recent decades, there has been a tremendous evolution in the traffic on the Internet and enterprises networks. Networks assisted since the beginning to two phenomena: on the one hand, the birth of a multitude of applications, each posing different requirements; on the other hand, the explosion of personal mobile networking, with an ever increasing demand of devices that require connectivity. These trends resulted in increased network complexity, leading to difficult management and high costs. At the same time, evolution in the Information Technology (IT) field led to the birth of cloud computing and growth of virtualization technologies, opening new opportunities not only for companies but for individuals (be it PC or mobile users), as well as Service and Infrastructure Providers. Emerging technologies such as Software Defined Networking (SDN) and Network Functions Virtualization (NFV) seems to be promising solutions to today’s network problems. Neither standardized solutions, nor how to properly combine their usage to achieve flexible and proactive control management have been discovered yet. This Ph.D. thesis focuses on the exploration of three plane of functionality in which software-defined (computer) networks can be divided: the data, the control and the management plane. In this thesis, we present insights on several aspects of network virtualization, starting from virtual network performance of cloud computing infrastructures, and introducing the Service Function Chaining (SFC) mechanism, discussing its analysis, design and implementation. In particular, the original contribution of this dissertation concerns (i) performance evaluation of the OpenStack cloud platform (the data plane); (ii) the design and implementation of a stateful SDN controller for dynamic SFC (the control plane); (iii) design, implementation and performance analysis of a proposed Intent-based approach for dynamic SFC (the management plane).