The Cost of Software-Defining Things: A Scalability Study of Software-Defined Sensor Networks

Software-Defined Networking is a promising paradigm for providing flexibility and programmability to computer networks. Our goal is to assess the performance of this paradigm applied to Wireless Sensor Networks. Previous evaluations are not complete, since they study small networks, do not explore crucial performance metrics, or solely examine light traffic conditions. For this, we execute simulations and a testbed experiment. The testbed shows Software-Defined Networking successfully operates in a real network. We study simulated networks up to 289 data-transmitting nodes, while assessing all the main networks metrics: data delivery, delay, control overhead, and energy consumption. We investigate important parameters for Software-Defined Wireless Sensor Networks, such as controller positioning, radio duty cycling, number of data sinks, and use of source routed control messages. The results indicate that Software-Defined Networking is feasible for Wireless Sensor Networks, presenting competitive data delivery ratio while saving energy in comparison to RPL, the Routing Protocol for Low-power and lossy networks.