I was wondering if there are any studies on the performance of the SparQL query service for either DBpedia or Wikidata. The query optimization page has some ideas, but I was looking at something more in-depth or formal. I am particularly interested in knowing about the limits, that is, how demanding does a query has to be for it to fail, etc…
Thank you in advance.
EDIT: I found this https://iccl.inf.tu-dresden.de/w/images/5/5a/Malyshev-et-al-Wikidata-SPARQL-ISWC-2018.pdf
@sanyides try this: http://www.semantic-web-journal.net/content/evaluation-metadata-representations-rdf-stores-0
A statement from the first study that really surprised me was:
“Figure 4 shows the resulting average times that it took to answer various percentiles of queries on the (busier) “eqiad” cluster. The top 50% of the queries are answered in less than 40ms on average, while the top 95% finish in an average of 440ms. The top 99% can reach peak latencies that are close to the 60s timeout, but on average also stay below 40s. At the same time, the total number of timeouts generated by BlazeGraph on all servers during this interval was below 100,000, which is less than 0.05% of the requests received.”
Only 0.05% of real-life requests end up in a timeout? Is it still a similar rate since its publishing in 2018?
@sanyides do you know wikidata.org ? It is a great project, they also have a mailing list, where you can ask Wikidata-specific questions. At DBpedia, we maintain much more than a Wikibase and a Blazegraph. So any general questions regarding knowledge engineering, NLP, AI and Semantic Web are welcome here and we are also qualified to answer DBpedia-specific questions. Check out the Virtuoso report: https://medium.com/virtuoso-blog/dbpedia-usage-report-as-of-2018-01-01-8cae1b81ca71
Also http://www.semantic-web-journal.net/content/evaluation-metadata-representations-rdf-stores-0 reproduces the first report you sent and @jfrey can answer questions.