CONSTRUCT queries performance on a Spark-based Big RDF triplestore

Despite their potential, CONSTRUCT queries have gained little attraction so far among data practitioners, vendors and researchers. In this paper, we first exhibit performance bottlenecks of existing triplestores for supporting CONSTRUCT queries over large knowledge graphs. Then, we describe a novel Spark-based architecture for big triplestores, called TESS, that we have designed and implemented to overcome the above limitations by using parallel computing. TESS ensures ACID properties that are required for a sound and complete implementation of CONSTRUCT-based forward-chaining rules reasoning.