An Evaluation of SPARQL Federation Engines Over Multiple Endpoints

Due to decentralized and linked architecture underlying Linking Data, running complex queries often require collecting data from multiple RDF datasets. The optimization of the runtime of such queries, called federated queries, is of central importance to ensure the scalability of Semantic-Web and Linked-Data-driven applications. This has motivated a considerable body of work on SPARQL query federation. However, previous evaluations of SPARQL query federation engines do not evaluate the performance of these engines pertaining to the different steps involved in the federated query processing. Consequently, it is difficult to pinpoint the components of the federation engines that need to be improved. This work presents an extended summary of the fine-grained evaluation of SPARQL endpoint federation systems performed in [13]. Beside query runtime as an evaluation criterion, we extend the scope of our performance evaluation by considering additional measures which are important but have not been paid much attention to in the previous studies. Our experimental outcomes lead to novel insights for improving current and future SPARQL federation systems.