Regresar

How to benchmark RDF engines and not die trying

Abstract:

Existing RDF engines have developed optimization techniques and efficient physical operators that speed up execution time. Additionally, some of these approaches have implemented structures to efficiently store and access RDF data, and have developed execution strategies able to reuse data previously stored in cache. In order to study the benefits of the techniques offered by each approach, particular datasets have to be considered as well as different benchmarks of queries. Queries must be of different complexity, number of patterns, selectivity and shape, while datasets have to be of different sizes and correlations between subjects, pbkp_redicates and objects. In this paper, we describe the benchmarks that we have developed to analyze the performance of existing RDF engines when creating and indexing RDF structures, as well as the quality of the optimization and execution strategies implemented by these engines.