Q-NiGHT: Adding QoS to Data Centric Storage in Non-Uniform Sensor Networks

Storage of sensed data in wireless sensor networks is essential when the sink node is unavailable due to failure and/or disconnections, but it can also provide efficient access to sensed data to multiple sink nodes. Recent approaches to data storage rely on Geographic Hash Tables for efficient data storage and retrieval. These approaches however do not support different QoS levels for different classes of data as the programmer has no control on the level of redundancy of data (and thus on data dependability). Moreover, they result in a great unbalance in the storage usage in each sensor, even when sensors are uniformly distributed.  This may cause serious data losses, waste energy and shorten the overall lifetime of the sensornet.  In this paper, we propose a novel protocol, Q-NiGHT, which (1) provides a direct control on the level of QoS in the data dependability, and (2) uses a strategy similar to the rejection method to build a hash function which scatters data approximately with the same distribution as sensors.  The benefits of QNiGHT are assessed through a detailed simulation experiment, also discussed in the paper. Results show its good performance on different sensors distributions on terms of both protocol costs and load balance between sensors.