Lessons Learned from Implementing Ad-hoc Multicast Routing in Sensor Networks

The mobile ad-hoc networking (MANET) community has proposed a wide range of protocols for unicast and multicast routing through mobile wireless devices. Many of these protocols have been studied only under simulation using simplistic radio models, and do not consider issues such as bandwidth or memory limitations. In contrast, the sensor network community has demanded solutions that work on hardware with limited resources, with a focus on routing through stationary nodes to a single base station. Still, several emerging sensor network applications involve mobile nodes with communication patterns requiring any-to-any routing topologies. We should be able to build upon the MANET work to implement these systems. However, translating these protocols into real implementations on resource-constrained sensor nodes raises a number of challenges. In this paper, we present the lessons learned from implementing one such protocol, Adaptive Demand-Driven Multicast Routing (ADMR), on CC2420-based motes using the TinyOS operating system. ADMR was chosen because it supports multicast communication, a critical requirement for many pervasive and mobile applications. To our knowledge, ours is the first non-simulated implementation of ADMR. Through extensive measurement on a 30-node sensor network testbed, we present the performance of TinyADMR under a wide range of conditions. We highlight the real-world impact of path selection metrics, radio asymmetry, protocol overhead, and limited routing table size.