Parallel Processing of Sensor Network Data Using Column-oriented Databases

Many wireless sensor network (WSN) applications require join of sensor data belonging to various sensor nodes. For join processing, it is important to minimize the communication cost since it is the main consumer of battery power. In this paper, we introduce a parallel join technique for sensor networks. A WSN consists of many independent sensor nodes and provides a natural platform for a shared-nothing architecture to carry out parallel processing. The proposed parallel join algorithm is based on sensor data that are stored in column-oriented databases. A column-oriented database store table data column-wise rather than row-wise as in traditional relational databases. The proposed algorithm is energy-efficient for two clear reasons. First, unlike relational databases, only relevant columns are shipped to the join region for final join processing. Second, parallel join processing of sensor data also improves performance. The performance analysis shows that the proposed algorithm outperforms join algorithms for sensor data that are based on relational databases.