SPARQL query processing with Apache Spark

The number and the size of linked open data graphs keep growing at a fast pace and confronts semantic RDF services with problems characterized as Big data. Distributed query processing is one of them and needs to be efficiently addressed with execution guaranteeing scalability, high availability and fault tolerance. RDF data management systems requiring these properties are rarely built from scratch but are rather designed on top of an existing engine. In this work, we consider the processing of SPARQL queries with the current state of the art cluster computing engine, namely Apache Spark. We propose and compare five different query processing approaches based on different join execution models and Spark components. A detailed experimentation on real-world and synthetic data sets promotes two new approaches tailored for the RDF data model which outperform (by a factor of up to 2.4 on query execution time compared to a state of the art distributed SPARQL processing engine) the other ones on all major query shapes, i.e., star, snowflake, chain and their composition.