Index Support for SPARQL

The Resource Description Framework (RDF) is the fundamental data model underlying the Semantic Web. Recently, SPARQL has been proposed as W3C standard for querying RDF repositories. As RDF is a graph-based data model, the core problem of evaluating SPARQL queries is matching query graph patterns against the data graph, which is computationally very expensive. We address this problem by indexing graph patterns. In the spirit of SQL, we assume that users select the patterns to be indexed. We formally define the problem of covering indexes, i.e., finding those indexes whose graph-patterns are contained in the query pattern, and derive formulas for estimating index selectivity. Finally, we study the problem of finding optimal sets of indexes for a given query. We believe that our framework is the first comprehensive suggestion for indexing RDF for SPARQL queries that goes beyond simple indexing of labels.