Efficient Query Processing for SPARQL Federations with Replicated Fragments

Abstract: Low reliability and availability of public SPARQL endpoints prevent real-world applications from exploiting all the potential of these querying infras-tructures. Fragmenting data on servers can improve data availability but degrades performance. Replicating fragments can offer new tradeoff between performance and availability. We propose FEDRA, a framework for querying Linked Data that takes advantage of client-side data replication, and performs a source selection algorithm that aims to reduce the number of selected public SPARQL endpoints, execution time, and intermediate results. FEDRA has been implemented on the state-of-the-art query engines ANAPSID and FedX, and empirically evaluated on a variety of real-world datasets.