RDF DATABASES – CASE STUDY AND PERFORMANCE EVALUATION

Abstract

The Resource Description Framework (RDF) data presentation model and the SPARQL query language have been the core of the semantic web technologies since the early 2000’s. In this article, we evaluate three RDF storage technologies. Our motivation is to find a storage solution that can be used to process “big data” RDF sets. Our method is based on measuring query response times with large samples (hundreds of thousands of RDF documents, millions of RDF statements). We find that all the proposed technologies provide much better performance than querying RDF data stored in files. However, with 300 000 documents, even with the fastest technology, an aggregation query still lasts more than 100 seconds in our environment. As a further performance improvement, we test the same data and queries with MongoDB, demonstrate its performance (10 seconds instead of 100) and scalability (up to 1000 000 documents). However, despite its benefits we must note that because of its data presentation and query limitations, MongoDB probably cannot serve as a generic storage for all kinds of RDF documents.