Efficient XML Data Management and Query Evaluation in Wireless Sensor Networks

With the advancing development in the field of microprocessor technology tiny wireless computing devices for sensory tasks denoted as wireless sensor nodes have been introduced. Large scale networks of these devices, denoted as wireless sensor networks, are deployed for long term location monitoring and typically collect large sets of sensor data over months or years. Managing sensor data in networks consisting of hardware resource limited sensor node devices brings up many challenges for database research. Moreover, integrating wireless sensor networks in heterogeneous networks is a complex task. A reason is the absence of a standardized data exchange format that is supported in all participating sub networks. In the last decade, XML has become the defacto standard for data exchange in the World Wide Web (WWW). The positive benefits of data exchangeability to support system and software heterogeneity on application level and easy WWW integration make XML an ideal data format for many other application and network scenarios like wireless sensor networks. Moreover, the usage of XML encourages using standardized techniques like SOAP to adapt the service oriented paradigm to sensor network engineering. Nevertheless, integrating XML usage in wireless sensor network data management is limited by the low hardware resources that require efficient XML data management strategies suitable to bridge the general resource gap. This dissertation introduces approaches for integrating efficient XML usage in wireless sensor networks. This includes the integration of XML in the engineering process, energy and memory efficient data management strategies, efficient solutions for XML data acquisition and general optimization strategies for handling XML queries and result messages in large scale sensor networks. In detail, this work introduces the programming framework XOBESensorNetwork which provides the direct use of XML in a sensor network programming language while ensuring stable and space, time and energy efficient programs handling XML data. To allow flexible XML data management on sensor node devices with strict hardware resource limitations, XOBESensorNetwork includes two separate strategies on integrating compressed XML data management in wireless sensor networks that both have been implemented and are running on today’s sensor node platforms. In this dissertation, it is shown how this compressed XML data can be further processed and how XPath queries can be evaluated dynamically. In an extended evaluation we compare the performance of both strategies concerning the memory and energy efficiency and show that both solutions have application domains and are fully applicable on today’s sensor node products. In summary, XOBESensorNetwork offers a complete XML solution for wireless sensor networks from application engineering to in-field data management. As part of the DFG project AESOP’s TALE, the presented XML data management solutions are the future basis for integrating SOAP support in wireless sensor networks. While the previous aspects cover the field of data management and application engineering, this dissertation also includes further optimizations in the field of communication. Generally, saving energy in wireless sensor networks is essential to extend the lifetime of in-field deployments. Previous research has shown that communication is generally the most energy consuming task and needs to be reduced in order to build resource-efficient long-term applications. This dissertation therefore additionally introduces optimizations for processing high amounts of unique queries by using a dynamic approximative caching scheme: DACS. In DACS, query results can be retrieved from caches that are placed nearer to the query source instead of sending queries deep into the network. The communication demand can be significantly reduced and the entire network lifetime is extended. To verify cache coherence in sensor networks with non-reliable communication channels, an approximative update policy is used. To localize the adequate cache adaptively, model-driven queries including a degree of demanded result quality can be defined. The entire logic is thereby processed by DACS and hidden to the user. The significant energy conservation is proven in evaluations that include real sensor node deployments.