Performance Evaluation of Controllers in Low-power IoT Networks

Abstract

Software-Defined Networking (SDN) enables network reconfiguration in response to dynamic application requirements. Recently researchers have tried extending SDN in the Internet of Things (IoT) networks, which required domain-specific customization in SDN architecture and protocols. Several architectures and designs have been proposed without adequate performance evaluation except for a couple: microSDN and SDN-WISE. microSDN architecture is built on a standard protocol stack, whereas SDN-WISE proposes a custom one. We first perform an extensive evaluation of these two architectures in terms of energy consumption, latency, and packet delivery ratio (PDR). The results confirm that microSDN can significantly reduce resource utilization while offering high PDR compared to SDN-WISE. Thus, we recommend microSDN as a potential architecture for low-power IoT networks. However, which SDN controller the chosen architecture can use? There is no standard evaluation bench-marking available. Thus, we evaluate the embedded (part of the IoT networks) microSDN and standard (external to the IoT networks) ONOS controllers on their throughput, delay, topology detection, and topology update time. We observe a trade-off between the scalability and performance between the embedded and external controllers. Thus, users can choose a controller based on their application demand, e.g., an embedded one for low traffic rate from a small number of IoT sources.