SPOVC: A scalable RDF store using horizontal partitioning and column oriented DBMS

Organizing and indexing RDF data for efficient evaluation of SPARQL queries has been attracting a lot of attention in the recent past. Most of the techniques proposed in this context leverage the existing RDBMS or column oriented DB technologies. In this paper, we propose an organization SPOVC that uses five indexes, namely, Subject, Predicate, Object, Value and Class, on top of any column oriented DB. The main techniques used by the proposed scheme are horizontal partitioning of the logical indices and special indices for values and classes. The SPOVC approach has the advantage of delivering better performance if the underlying column store technology improves. The proposed approach is conceptually much simpler than the state-of-the-art native-storage based proposals and roughly gives the same performance. Our proposal extends an existing approach, SW-Store, that uses column oriented DBs and vertical partitioning and obtains a two/three fold performance improvement. In addition, the proposed system is the only system that can effectively tackle SPARQL queries with filter patterns having range conditions and regular expressions. © 2012 ACM.