NarrowCast: A New Link-Layer Primitive for Gossip-Based Sensornet Protocols

Although gossiping protocols for wireless sensor networks (sensornets) excel at minimizing the number of generated packets, they leave room for improvement when it comes to the end-to-end performance, namely energy efficiency. As a step in remedying this situation, we propose NarrowCast: a new primitive that can be provided by asynchronous duty-cycling link layers as a substitute for broadcasting for gossiping protocols. The principal idea behind the NarrowCast primitive is to allow a sensor node to transmit to a fraction of its neighbors, which enables controlling energy expenditures and reliability. We discuss methods of approximating the primitive in practice and integrating it with gossiping protocols. We also evaluate implementations of the approximations with Trickle, a state-of-the-art gossiping protocol, and X-MAC, a popular link layer based on low-power listening. The results show that—without sacrificing reliability—gossiping using even the simplest approximations of NarrowCast can considerably outperform gossiping based on broadcasting in energy efficiency.