Is QLever a standalone alternative for a triplestore like Jena Fuseki?

[Stiksels]

Hi, we are currently investigating rdf storage and query alternatives for our current setup of Jena Fuseki with TDB. Everything we've heard about and experienced with QLever makes this a very interesting solution.

My question is mainly; like Jena needs the combination of Fuseki and TDB, does the QLever SPARQL server still need a triplestore for rdf persistence?

Kind regards,
Stan

[joka921]

Hello Stan,
QLever is a standalone solution that handles the indexing + storing of the RDF data as well as the querying of the data via RDF.
I think it's best for you to try our Quickstart guide on the repositories main page using the qlever control script with one of our smaller example knowledge graphs. You can then try to load your own data into QLever, let us know if you run into trouble, and also, if it works as expected. What kind of data are you working on (number of triples, domain, etc.), and is it publicly available? We are always looking for interesting data sets to test our engine.

Best regards
Johannes

[Stiksels]

Hi @joka921 ,
Thanks for your reply. We are working on a interoperable data exchange platform for cultural participation data in Flanders (Events, Locations, Organizers, Performers, Participants, Participations...). It follows the data mesh principles and has +20 data products at the moment, with 4.5mio subjects (+225mio triples). For the foreseable future we guestimate +40 data products; 150mio subjects (8bn triples). We'll also need to aggregate datasets and perform reasoning and entity resolution.

[Stiksels]

To really test QLever as a solution, we'll need the named graph support that is currently still in development, as most of the datasets use named graphs to describe domain entities

[tuukka]

most of the datasets use named graphs to describe domain entities

Does this mean the number of named graphs will be much higher than the number of data products? I'm thinking about what kinds of experiments and test cases you could already make before named graphs become available in QLever.

[Stiksels]

Yes, the subject count is actually the graph count over all dataproducts; so they are more like named entity graphs, with an average of 60 triples per graph

[hannahbast]

@Stiksels Do you have a link to your dataset or a subset of it? QLever will have support for named graphs soon (there already is a PR). Until then, there are workarounds.

[Stiksels]

@hannahbast I would be happy to provide you with a data dump but am currently in holiday mode without access to pc until the 22nd of July 😳

[Stiksels]

@hannahbast would it be best to wait for the PR or try a workaround? I can provide you a compressed nquads datadump file for one dataproduct (2.3mio quads), but that should only be for academic use. Can I contact you directly at your public email [email protected] ?