Twig’n Join: Progressive Query Processing of Multiple XML Streams

Abstract

We propose a practical approach to the progressive processing of (FWR) XQuery queries on multiple XML streams, called Twig’n Join (or TnJ). The query is decomposed into a query plan combining several twig queries on the individual streams, followed by a multi-way join and a final twig query. The processing is itself accordingly decomposed into three pipelined stages progressively producing streams of XML fragments. Twig’n Join combines the advantages of the recently proposed TwigM algorithm and our previous work on relational result-rate based progressive joins. In addition, we introduce a novel dynamic probing technique, called Result-Oriented Probing (ROP), which determines an optimal probing sequence for the multi-way join. This significantly reduces the amount of redundant probing for results. We comparatively evaluate the performance of Twig’n Join using both synthetic and real-life data from standard XML query processing benchmarks. We show that Twig’n Join is indeed effective and efficient for processing multiple XML streams.