Press the Cancel Button! A Performance Evaluation of Scalable In-Network Data Aggregation

We perform real-world tests of the performance of medium access control protocol schemes for scalable data aggregation in sensor networks. Specifically, we evaluate the performance of a Listen-and-Suppress Carrier Sense Multiple Access (LAS-CSMA) scheme on the duplicate-insensitive exemplary monotonic aggregates MAX and MIN. These schemes reduce power consumption, network bandwidth usage and delays by suppressing node packet transmissions that are proven to be unnecessary in the query response. This is possible when nodes listen to the transmissions of other nodes as they respond. Scalability tests were performed for 8 networks of various sizes, the largest having 24 Crossbow IRIS wireless mote modules as data nodes. We found that for the largest networks, packet transmissions, receptions, and response delays were reduced 44%, 65%, and 38% respectively. Packet transmissions and response delays were found to scale nearly logarithmically with respect to the size of the network, which resembles predictions by recent theoretical results. However, the consumption of energy and network resources is further reduced in platforms with better software implementations and hardware support for the TinyOS Active Messages interface. It is found that the scalability and performance of LAS-CSMA depends strongly on how quickly nodes can cancel a pending packet transmission based on a calculation with data from an incoming packet.