| Abstract |
RDF store systems are the backbones of the applications driven by RDF data. There is a number of benchmark systems available to assess performances of the RDF stores. The performance of the systems may vary from one application domain to another. This imposes a need for application benchmarks that closely mimic real-life scenarios, e.g., by using real-life data sets from the domain of interest and use case motivated queries.
In this talk I will introduce the ParlBench benchmark that simulates a large scale electronic publishing scenario. The goal of the benchmark is to assist developers of e-publishing applications in choosing RDF store systems that are more suitable for analytical query processing. The benchmark offers large collections of real-life data sets including Dutch parliamentary proceedings and a set of analytical queries developed on top of this data. In the first part of the talk I will focus on the benchmark data sets which were obtained by transforming the PoliticalMashup proceedings data set available in XML (which is also used in the PoliMedia project of Laura Hollink). I will discuss the design choices that were made when developing the underlying RDF model of the PoliticalMashup data. In the second part of the talk, we evaluate the benchmark on the Virtuoso OSE RDF store, and discuss the results obtained. A preliminary version of the results was submitted to the Workshop on Benchmarking RDF Systems [1] co-located with ESWC 2013. |