IrisNet: An Architecture for Compute-Intensive Wide-Area Sensor Network Services

Previous work on sensor networks has targeted ad hoc wireless networks of closely-colocated, resourceconstrained scalar sensor motes. Such work has overlooked richer sensor types such as webcams and microphones, which are typically attached to Internetconnected machines with significant computing power and storage. In this paper, we describe IrisNet (Internetscale Resource-Intensive Sensor Network services), an architecture for wide-area sensor networks based on these more capable sensing nodes. IrisNet provides a common, scalable software infrastructure that enables the flexible creation of sensor-based Internet services. It dramatically reduces network bandwidth utilization through the use of senselets, binary code fragments that perform intensive data filtering at the sensing nodes, leveraging the available processing power and memory. IrisNet employs a two-tier hierarchy of sensing nodes and query processing nodes. Key features of IrisNet include flexible data partitioning, efficient and protected sharing of sensor nodes, low-latency queries, partial match caching, query-specified freshness tolerances, and monitoring and logging support. This paper reports on experiments with a working IrisNet prototype running a parking space finder service, demonstrating the effectiveness of IrisNet’s features in achieving scalability and reducing query response times.