High-Performance XML Twig Filtering using GPUs

Current state of the art in information dissemination comprises of publishers broadcasting XML-coded documents, in turn selectively forwarded to interested subscribers. The deployment of XML at the heart of this setup greatly increases the expressive power of the profiles listed by subscribers, using the XPath language. On the other hand, with great expressive power comes great performance responsibility: it is becoming harder for the matching infrastructure to keep up with the high volumes of data and users. Traditionally, general purpose computing platforms have generally been favored over customized computational setups, due to the simplified usability and significant reduction of development time. The sequential nature of these general purpose computers however limits their performance scalability. In this work, we propose the implementation of the filtering infrastructure using the massively parallel Graphical Processing Units (GPUs). We consider the holistic (no post-processing) evaluation of thousands of complex twig-style XPath queries in a streaming (single-pass) fashion, resulting in a speedup over CPUs up to 9x in the single-document case and up to 4x for large batches of documents. A thorough set of experiments is provided, detailing the varying effects of several factors on the CPU and GPU filtering platforms.