CONTROL PLANE DEPENDABILITY IN SOFTWARE DEFINED NETWORKS

Abstract

The separation of control plane and data plane in Software Defined Networks (SDN) brings flexibility and programmability to network management. Controllers in the control plane manage the flow tables in switches through protocols such as Openflow and can be seen as the “brain” of a network. Thus, it is imperative that the control plane is “dependable” for smooth network operation. Dependability implies that the control plane is available, reliable and trustworthy. This thesis examines issues related to these aspects and proposes solutions to improve control plane dependability. Control traffic engineering policies are proposed to improve availability of controllers and manage their performance in terms of response time. Quality of Controller (QoC) metric (considering controller reliability and response time) and switch-controller mapping solutions to maximize QoC are proposed to improve reliability. Game-theoretic and machine learning based solutions are proposed to identify potentially malicious controllers and ensure trustworthiness of the control plane.