An Efficient Processing of Join Queries for Sensor Networks Using Column-Oriented Databases

Abstract

Recently, the sensor network area is gaining attention both in the industry and academia. Many applications of sensor network such as vehicle tracking and environmental monitoring require joining sensor data scattered over the network. The main performance criterion for queries in a sensor network is to minimize the battery power consumption in each sensor node. Hence, reducing the communication cost of shipping data among sensor nodes is important since it is the main consumer of battery power. In this paper, we propose a technique for join queries in a sensor network that minimizes communication cost. For storage of sensor data, we use a column-oriented database that stores data on disk (or in memory) column-by-column unlike traditional database that store data in rows. The justification for using a column-oriented database technique is not to ship those data columns that do not participate in the actual join. We compare our algorithm with existing join algorithms for sensor networks that are based on traditional row-oriented databases. The performance analysis show that our proposed algorithm based on column-oriented databases outperforms existing algorithms in processing binary equi-join (BEJ) queries for sensor networks.