Constructing Large-Scale Semantic Web Indices for the Six RDF Collation Orders

The Semantic Web community collects masses of valuable and publicly available RDF data in order to drive the success story of the Semantic Web. Efficient processing of these datasets requires their indexing. Semantic Web indices make use of the simple data model of RDF: The basic concept of RDF is the triple, which hence has only 6 different collation orders. On the one hand having 6 collation orders indexed fast merge joins (consuming the sorted input of the indices) can be applied as much as possible during query processing. On the other hand constructing the indices for 6 different collation orders is very time-consuming for large-scale datasets. Hence the focus of this paper is the efficient Semantic Web index construction for large-scale datasets on today's multi-core computers. We complete our discussion with a comprehensive performance evaluation, where our approach efficiently constructs the indices of over 1 billion triples of real world data.