Query execution for RDF data on row and column store

Abstract

This paper shows experimental comparison between various data storage techniques to manage RDF data. The work represents evaluation of query performance in terms of query execution time and data scalability, using row and column store for various data storage techniques. To demonstrate these ideas FOAF (Friend Of A Friend) data is used. The paper contributes experimental and analytical study for application of partitioning techniques on FOAF data which makes queries 168 times faster compared to traditional triples table. Materialized views over vertically partitioned data show an additional 8 times improvement in query performance against partitioned data for the frequently occurring queries. Vertical partitioning is executed on column store also, and as FOAF data size scales, an order of magnitude improved performance is observed over row store execution.