Should We Be Afraid of Querying Billions of Triples in a Graph-Based Centralized System?

Data representation facilities offered by RDF (Resource Description Framework) have made it very popular. It is now considered as a standard in several fields (Web, Biology,...). Indeed, by lightening the notion of schema, RDF allows a flexibility in the representation of data. This popularity has given rise to large datasets and has consequently led to the need for efficient processing of these data. In this paper, we propose a novel approach that we name QDAG (Querying Data as Graphs) allowing query processing on RDF data. We propose to combine RDF graph exploration with physical fragmentation of triples. Graph exploration makes possible to exploit the structure of the graph and its semantics while the fragmentation allows to group the nodes of the graph having the same properties. Compared to the state of the art (i.e., gStore, RDF3X, Virtuoso), our approach offers a compromise between efficient query processing and scalability. In this regard, we conducted an experimental study using real and synthetic datasets to validate our approach with respect to scalability and performance.