Aggregating sensor data from overlapping multi-hop network neighborhoods: Push or pull?

Network neighborhoods are a key communication abstraction in sensor networks, allowing sensor nodes to collect and aggregate sensor data from nearby other nodes. In many applications, multi-hop neighborhoods of several nodes overlap, such that nodes participate in many neighborhoods, having to contribute their data items to all containing neighborhoods. We consider two orthogonal approaches to efficiently support this data aggregation problem. A push-based approach, where each node floods its data item in a multi-hop neighborhood, and a pull-based approach, where each node collects data from nodes in a multi-hop network neighborhood using a spanning tree. Our goal is to identify situations where one approach outperforms the other. For this, we implement these protocols in TOSSIM, study overhead and yield as a function of the fraction of nodes in the network that perform data aggregation over a multi-hop neighborhood, and report our findings.