Data-stream processing has become critical in numerous present-day applications. Example stream applications that use Extensible Markup Language (XML) for representing the data include selective dissemination of information such as personalized news delivery, XML message brokers for Business to Business (B2B) message-oriented middleware servers. This project explores a novel paradigm for XML stream processing that integrates token-based automata computation and tuple-based algebraic computation into one uniform framework. The resulting Raindrop architecture supports the flexible exploitation of both automata and algebra computation paradigms. This, among other advantages, enables the application of mature algebraic optimization techniques such as query plan rewriting and cost estimation within this new context. Furthermore, this approach enables pushing operators in and out of the automata, based on data statistics. The Raindrop framework incorporates additional advanced features, such as schema-based optimization, multiple execution modes and run-time physical plan reformulation.
A long-reaching impact is expected in many real-world applications that are Internet centric, as these can benefit from the resulting XML-based stream monitoring technology. Broader impacts include the development and dissemination of the software systems developed via the project's Web site (http://davis.wpi.edu/~dsrg/raindrop), and the integration of the PIs' research activities with undergraduate and graduate teaching in line with the project-based spirit of education at WPI.
