A GeoKnow Benchmarking laboratory has been setup for comparison of benchmark results over the duration of the GeoKnow project. In it current form this takes for the form of original the LOD2 project GeoBench program which has been taken over and adopted for use in the GeoKnow project and available from the GeobenchLab GIT repo.
The current improvements in the GeoBench program are primarily related to the expansion of the benchmark, in order to make it employable not only to RDF data, but to relational data as well. This will open opportunities for a performance comparison between RDF and relational spatial data management systems.
Below are some comparison results using the planet wide Open Street Map (OSM) datasets hosted in both Virtuoso and PostGIS.
In summary, the result demonstrate that Virtuoso in both SQL and SPARQL outperformed PostGIS by significant factor. Specifically, all the queries in the power run were executed 33 times slower in PostGIS than in Virtuoso SQL (single server). If we compare PostGIS with Virtuoso SPARQL, the factor will be even greater: 131 for low zoom level queries, 23 for high zoom level queries, and 113 in total. If we correlate Virtuoso SPARQL and SQL (single server), we will conclude that the relational version is slower almost 4 times on low zoom level queries, while it is faster 23% on high zoom level. In total, SQL version is slower more than 3 times.
The full descriptions and results of these benchmarks can be found in the GeoKnow D1.3.2 - Continuous Report on Performance Evaluation deliverable.