Distributed Evaluation of XPath Axes Queries over Large XML Documents Stored in MapReduce Clusters

The MR (MapReduce) framework, a programming model for parallel computation over data stored in a cluster of commodity computers, established itself as one of the leading solutions for Big Data processing. This framework is also being used like a query language in many database systems, because it can process data stored in various unstructured, semi-structured, and structured formats. Nevertheless, the MR framework can be used for XML data processing too, it does not allow to write queries in a declarative manner, like XPath or XQuery. To overcome this problem, we propose a system that enables to query XML data with XPath, but it evaluates the queries in parallel using the MR framework. First, we introduce a persistent storage that maps XML data into a wide-column store. The proposed mapping enables efficient and distributed data processing. Secondly, we describe a query processor translating an XPath language subset to MR jobs. Finally, we present tests and their results showing the scalability of our system.