RDFox (http://www.cs.ox.ac.uk/isg/tools/RDFox/) is a state of the art RDF management system that is currently being developed at the University of Oxford. The core of the system runs in main memory, and it supports reasoning over RDF datasets using datalog with stratified negation and aggregation. The reasoning algorithm is based on materialisation: datalog rules are applied iteratively until all relevant consequences have been derived. This process is computationally very intensive, so RDFox uses a new approach to parallelise computation in multi-core systems. The approach comprises two components: a new algorithm for seminaive datalog evaluation, and new RDF indexing data structures that efficiently supports concurrent queries and updates. In this talk, I will present an overview of the parallelisation approach, and I will particularly focus on lock-free updates of data structure.
In order to support data loads that cannot fit into main memory of a single server, RDFox provides a distributed solution that supports answering queries over a data set distributed in a shared-nothing cluster. Communication between cluster nodes is a primary source of overhead in such systems, so RDFox attempts to minimise communication using a new query answering approach based on dynamic data exchange. RDFox also uses a novel algorithm for data partitioning based on graph partitioning whose goal is to minimise the need for communication. In this talk, I will present the main ideas behind the dynamic exchange and graph partitioning.
Boris Motik is a Professor of Computer Science at the University of Oxford. He got his PhD from the University of Karlsruhe under supervision of Prof Rudi Studer, after which he joined the group of Prof Ian Horrocks at the University of Manchester and subsequently moved to Oxford. His interest include knowledge representation formalisms, reasoning algorithms, databases, and practical applications of semantic technologies. He is the main author of five documents of the OWL 2 ontology language standard. Finally, he developed and applied in practice several reasoning systems such as RDF/datalog systems KAON and RDFox, and OWL/description logic reasoners KAON2 and HermiT.