A Cost-based Optimizer for SPARQL Queries

The cost of answering a query against an ontology is affected by at least three elements: the size of the ontology, the strategy followed to combine the data, and the order or plan in which data is processed. In the context of the Semantic Web, very large ontologies have been defined; therefore, techniques to identify efficient evaluation strategies are needed. We propose cost-based optimization techniques for SPARQL queries. In our approach, ontologies are modeled as a deductive database. The extensional database is comprised of meta-level predicates that represent the information explicitly modeled by the ontology; for each RDFS built-in vocabulary term, we define a meta-level predicate (e.g., subClassOf). The intensional database corresponds to the deductive rules that implement the semantics of the vocabulary terms (e.g., the transitive properties of the subClassOf term). Currently, we have developed the following techniques: