LocRDF: An Ontology-Aware Key-Value Store for Massive RDF Data

With the rapid development of the Semantic Web, the scale of RDF graphs surges. To describe ontology information, RDFs and OWL are endorsed by W3C, which further enhances the expressiveness of RDF graphs. A great challenge of managing RDF graphs is how to store massive data and efficiently reason ontology information at query time. There are two main issues with the existing RDF graph storage systems: 1) the relational data model is mainly used as the underlying storage architecture, which not only leads to exceeding the storage capacity, but also may incur high overhead while performing complex queries or multi-join queries; 2) the ontology reasoning module is either relatively independent of storage layer or used as an upper-layer application of storage and query system, causing redundancy and inefficiency in query. To address these issues, we present LocRDF, a novel storage system for RDF graphs via key-value store supporting ontology reasoning. LocRDF integrates ontology information into the underlying storage scheme with the application of a fixed-length interval encoding, promoting the efficiency of ontology reasoning at runtime. Experimental results on LUBM datasets show that extended ontology reasoning on large-scale RDF graphs scarcely affects query performance which is even significantly better than the existing state-of-the-art RDF query engines.