Energy efficient query-specific in-network query processing in sensor networks

Abstract

In sensor networks, summarized data are often more interested than raw, detailed data. Since sensor nodes have limited power, it is important to process aggregation queries energy efficiently to prolong the lifetime of sensor networks. In this dissertation, we are concerned with energy efficient aggregation query processing in sensor networks.

In-network processing of aggregation queries is a widely accepted technique to reduce energy consumption for wireless communication, which is a main source of energy consumption in sensor nodes. The main idea of in-network processing is that aggregates are processed in the network as data flow through the sensor nodes. Unnecessary data are discarded and relevant sensor readings are combined into a more compact form whenever possible.

How to deliver sensor readings to the base station, where the user requests a query and receive the answer of the query, is important for energy efficient in-network processing. Since an efficient routing structure for in-network processing may be different query by query, a routing method should consider query-specific cost reduction for energy efficient in-network processing of aggregation queries. Most of the existing routing methods, however, do not consider queries when deciding how to route sensor readings. Although there is some work that considers query-specific cost reduction, the applicability of their work is limited. In other words, their methods are designed either for specific queries or for particular sensor network environments.

In this dissertation, we propose a query-specific routing approach to energy efficient in-network processing of aggregation queries in sensor networks. Aggregation queries either have grouping conditions or not. For aggregation queries without grouping, any two sensor readings can be merged into a small partial aggregate; on the other hand, for grouped aggregation queries, only those sensor readings in the same group can be merged. Because of this ...