Benchmarking over a Semantic Repository

Abstract

In recent times, tremendous amount of work has been done to explore alternative methods to store and query Resource Description Framework (RDF) data because scalability and fast query performance are necessary requirements for real world semantic applications built on top of RDF stores. Recent works like Hexastore, Matrix Bit Loaded and Vertical partitioning approach deals with storing, querying and inferencing RDF data. SPARQL is a W3C standard query language used to query RDF data. It is not possible to express subsumption bound queries in SPARQL as the underlying data model does not provide support for type information in the subject and the object. In this paper we propose a new data model which is an extension of RDF data model to support the type information and on this extended data model we propose a new query language, PredQL (Predicate Query Language). We have used the Lehigh University benchmark (LUBM), to evaluate the performance, reasoning capabilities and storage mechanisms of our Semantic Repository. We have demonstrated our results against one billion triples. Our results show that our method of storing RDF data supports all types of queries with higher scalability and faster query performance.