Sensor networks can be viewed as large distributed databases, and SQL-like high-level declarative languages can be used for data and information retrieval. Energy constraints make optimizing query processing particularly important. This paper addresses for the first time, multiroot, multi-query optimization for long duration aggregation queries. The paper formulates three algorithms - naive algorithm (NMQ), which does not exploit any query result sharing, and two proposed new algorithms: an optimal algorithm (OMQ) and a heuristic (zone-based) algorithm (ZMQ). The heuristic algorithm is based on sharing the partially aggregated results of pre-configured geographic regions and exploits the novel idea of applying a grouping technique by using the location attribute of sensor nodes as the grouping criterion. Extensive simulations indicate that the proposed algorithms provide significant energy savings under a wide range of sensor network deployments and query region options.
Zhiguo Zhang, Ajay D. Kshemkalyani, Sol M. Shatz