Usage-Centric Benchmarking of RDF Triple Stores

A central component in many applications is the under-lying data management layer. In Data-Web applications, the central component of this layer is the triple store. It is thus evident that finding the most adequate store for the application to develop is of crucial importance for individual projects as well as for data integration on the Data Web in general. In this paper, we propose a generic benchmark creation procedure for SPARQL, which we apply to the DBpedia knowledge base. In con-trast to previous approaches, our benchmark is based on queries that were actually issued by humans and ap-plications against existing RDF data not resembling a relational schema. In addition, our approach does not only take the query string but also the features of the queries into consideration during the benchmark gen-eration process. Our generic procedure for benchmark creation is based on query-log mining, SPARQL feature analysis and clustering. After presenting the method un-derlying our benchmark generation algorithm, we use the generated benchmark to compare the popular triple store implementations Virtuoso, Sesame, Jena-TDB, and BigOWLIM. 1