Abstract

Software Defined Networking (SDN) is a new networking paradigm that facilitates a centralized system of computer networks by decoupling the control and data plane from each other, where a controller maintains the management of a global view of the network. SDN architectures can provide programmatic interfaces in communication networks that significantly simplify network management. Hence, the controllability and manageability of a network can be improved. On the one hand, the placement of controllers can significantly impact network performance in terms of controller responsiveness. On the other hand, SDN offers the ability to have controllers distributed over the network to solve the single point of failure problem at the control plane, increasing scalability and flexibility. However, there are some inevitable problems for such networks, especially for controller-related problems. For instance, scalability, reliability, and controller availability are some of the hottest aspects of SDN. More precisely, failure of the controllers themselves may lead to the impact of these aspects and the collapse of the network performance.
Despite the issues mentioned above, the controller placement challenges must be appropriately addressed to take advantage of the SDN. The connections between the controller (control plane) and the switches (data plane) in SDN are established by either an in-band or an out-of-band control mechanism. New challenges still arise regardin the connection availability and provide more protection for the connection between the data and control planes. A disconnection between the two planes could result in performance degradation. Although the SDN offers the advantage of an environment of multiple distributed controllers, yet the intercommunication factor between these controllers is still a key challenge. This thesis investigates the issues mentioned above and organizes them into four stages.
First, dealing with the controller placement problem as the most crucial concern in SDN, via exploiting the independent dominating set approach to ensure a distribution of controllers with lowest response times. We propose a new node degree-based algorithm named High Degree with Independent Dominating Set (HDIDS) for the controller placement problem in the SDN networks. HDIDS is composed of two phases to deal with controller placement: (1) determining candidate controller instances by selecting those nodes with the highest degree; and (2) partitioning the network into
multiple domains, one controller per domain.
To further improve network performance, reliability, and survivability, one solution is to deploy backup controllers to satisfy the quality of service requirements. In this regard, as a second step, we enhance the controller placement approach by designing a reliable and survivable controller placement strategy. This strategy relies on the efficient deployment of backup controllers by constructing virtual backup domains set(s) to ensure the durability and resilience of network control management. The approach design is called a Survivable Backup Controller Placement approach.
Furthermore, to achieve reliable control traffic between data and control planes in an in-band control network, as a third stage, we design and implement an In-band Control Protection Module that finds a set of ideal paths for the control channel under the failure conditions. The proposed protection mechanism protects as much control traffic as possible.
Finally, we present a practical approach for the controller placement problem in software defined networks aiming to minimize the inter-controller communication delay time and the delay time between controller and switches. The principal concept employed in this approach is the Connected Dominating Set. Further, we present an algorithm using the Minimum Connected Dominating Set, which minimizes the delay time between the distributed SDN controllers.