A Framework for Efficient Query Processing in Wireless Sensor Networks

Abstract

Though Wireless Sensor Networks (WSN) have invaded every walk of modern life due to low setup time and cost yet Research and Development (R&D) for further reduction in sensor costs and energy consumption is still a high priority area for research community. Major chunk of energy consumption of WSN is in communication activity therefore most of the research has targeted the aspect of reduction in energy consumption during communication of data. As a result of research efficient hardware and numerous network protocol stacks along with data reduction techniques have been developed. In a WSN, reduction of data is possible at two stages: „In-network‟ and at „Base Station‟ (BS). Due to limited processing capacity of sensor nodes, data processing in network is possible to a limited extent only whereas, at the BS the room for manipulation is much more. In this work processing at both the stages In-network aggregation and BS optimization has been used to bring about reduction in data to be communicated. The proposed framework:“Compression At inpuT with Multi query Optimization at Sink” (CATMOS) is built upon query type data extraction technique and makes use of both Innetwork optimization and Base Station optimization for efficient Query Processing. In the work, element of Network management has been employed by using In-network compression. Query optimization at BS, has been achieved by merging new queries with already running queries resulting into fewer new queries. Query language syntax has also been modified in which single syllable Static variables have been used in place of repeated phrases. The framework has been designed to benefit any WSN irrespective of application being served by it. The work has adopted a simple generalized approach for efficient operation of WSN, therefore, it is different, from most of the researches which has used highly technical or specialized means and have intended to be application specific. This proposed generalized approach is applicable to WSN of any size or topology. It isolates the user from the lower level details of network configuration or protocols used. In „In-network aggregation‟ the data has been compressed at the node level. The main objective has been to reduce the data size by eliminating spatio-temporal redundancies. Data fidelity has been given importance therefore quantitative reduction in data has been allowed without any compromise in quality of data. Reduced data to be communicated has produced desired result of less consumption of sensor node energy. Three compression algorithms i.e., Huffman, LZW and Deflate compression algorithms have been examined on a Sun Small Programmable Object Technology (SunSPOT) network. Possible data reduction through each algorithm has been tested on a simulator „Solarium‟. For BS optimization, a frugal approach has been adopted. In case of any new query, data available in response to already running queries within the network has been scrutinized to check whether the reply to the newly injected queries can be generated from the data or not. If at all, query has to be sent to the network, it has been merged with the running queries. A novel algorithm for rewriting queries after merging two or more queries has been proposed. New queries are rewritten along with the already running queries within the network. Only those mergers have been chosen for action, which provide actual Gains as verified through the algorithm. The merged queries have been given the name „Synthetic Queries‟. The proposed algorithm sees to it that the chosen synthetic queries achieve all the desired objectives. In addition to this query language syntax modification by using single syllable static variable has also been used to achieve additional Gains. Simulator results of compression and query optimization have been encouraging. More than 10% Gain has been possible through use of these simple compression algorithms only.