Clustering Based Distributed Controller Placement in Software Defined Wireless Sensor Networks

Abstract

In traditional network, each network device has control plane as well as data plane. newlineThis makes the devices bulky, difficult to manage, slow, and costlier. Software Defined newlineNetwork (SDN) makes the devices lighter by separating control plane and data plane newlineto make the network programmable. The control plane is hosted on a controller. The newlinecontroller performs the task of constructing forwarding rules. These rules are used newlineby switches in the data plane for movement of data. However, in a large network, a newlinesingle centralized controller will increase latency causing performance degradation. This newlinemandates using multiple controllers to optimize the network performance. newlineAdvancements in smart sensors have caused wide spread deployment of Wireless Sen- newlinesor Networks (WSNs) and IoT applications. WSNs require efficient data-transmission newlineand routing strategies. Finding routes in these networks is expensive in terms of energy newlineconsumption and requires more computing power within each node. This has encouraged newlinedeploying software-defined paradigms in WSNs. newlineSDWSN employs a control server and multiple Control Nodes (CNs). A base sta- newlinetion, termed as control server (CS), computes routes, and CNs are selected amongst the newlineSensor Nodes (SNs). Optimizing the number of CNs and finding optimal location poses newlinemany challenges. Exploration of these challenges in multi-sink networks require different newlineapproach. Scalability of the approach is also an under studied area. Controller Selection newlineproblem has been attempted through PITS and NWPSO. These approaches suffer from newlineimbalanced loading of controllers and link failure. In this thesis, we present SFO-CS newlineaimed at finding an optimal number of controllers co-hosted on the sensor nodes. The newlinemulti-objective controller selection and placement formulation optimizes energy con- newlinesumption through communication between SNs, CNs, and CS. The fitness function is newlineevaluated through energy and Euclidean distance parameters. Based on these param- newlineeters, the proposed SFO-CS selects an optimal number of controllers.