Cost-aware monitoring of network-wide aggregates in wireless sensor networks

Motivated by applications of wireless sensor networks, there have been growing interests in monitoring large scale distributed systems. In these applications, we usually wish to monitor global system conditions defined as a function of network measurements. In this paper, we study optimal strategies for reactive monitoring to be employed for monitoring network-wide aggregates of sensor nodes’ measurements. Our primary concern in adopting such a monitoring mechanism is to reduce the communication cost which is the dominant factor of energy drain in wireless sensor networks. To adapt the structure of monitoring mechanism to the statistics of nodes’ measurements, we devise a simple yet efficient algorithm that is appropriate for a class of distribution functions. Towards this, we consider a sigmoid approximation of the CDF of the underlying event and cast the underlying design problem as a convex optimization problem. This allows us to propose an algorithm to set monitoring parameters in accordance to the statistics of the events measured by spatially scattered sensor nodes. Through simulation experiments, we illustrate that the proposed algorithm, referred to as SATA, can significantly reduce the communication overhead of the monitoring mechanism in sensor networks. Our results show that compared to heuristic methods, the cost of monitoring mechanism can be reduced significantly.