Optimizing SPARQL-to-SQL Rewriting

Abstract

The vast majority of the structured data of our age is stored in relational databases. In order to link and integrate this data on the Web, it is of paramount importance to map relational data to the RDF data model and make Linked Data interfaces to the data available. We can distinguish two main approaches: First, the database can be transformed into RDF row by row and the resulting knowledge base can be exposed using a triple store. Second, an RDB2RDF mapper performs SPARQL-to-SQL rewriting and thus exposes a virtual RDF graph based on the relational database. The key challenge of such a SPARQL-to-SQL rewriting is to create a SQL query which can be efficiently executed by the optimizer of the underlying relational database. In this article we discuss and evaluate the impact of different optimizations on query execution time using SparqlMap, a R2RML compliant SPARQL-to-SQL rewriter and compare the performance with state-of-the-art systems.