SPARQL Benchmarking with Automatically Generated OLAP Queries

The growing use of data analytics on Linked Data requires SPARQL engines to efficiently execute Online Analytical Processing (OLAP) queries. While SPARQL 1.1 provides appropriate basic constructs, corresponding optimization of SPARQL engines is still in its infancy and further development lacks benchmarks that mimick the data distributions found in Link Data. In fact, existing work on OLAP benchmarking for SPARQL has usually adopted queries and data from relational databases, which may not well represent Linked Data. We map typical OLAP operations to SPARQL and propose a tool named ASPG to automatically generate OLAP queries from real-world Linked Data, which can be used to construct analytic benchmarks for SPARQL engines. We present such a benchmark called DBOB as an example which consists of queries generated using DBpedia. We apply this benchmark to a replication of DBpedia and present the result.