If you are using TDB, and since Fuseki 2.4.0, for in memory, graph is
treated just as a column like subject or predicate or object. It does
not make much difference; some more indexing (and even then if necessary
you can tune it).
Default union graph works and does not care much either.
You are right the UI is skewed to small number of graphs, but that is a
UI and admin issue.
The raw storage and query does not really see much difference. Default
union graph works and does not care much either.
Andy
On 17/01/17 14:16, Daniel Hernandez wrote:
I do a benchmark with different models with the same data. One model had
more than 65 millions of named graphs. Fuseki had no problems with named
compared with storing the same data in a single graph. You can see the
results of the benchmark in:
https://users.dcc.uchile.cl/~dhernand/research/ssws-2015-reifying.pdf
Cheers,
Daniel
---- On mar, 17 ene 2017 06:22:11 -0300 Nikolaos Beredimas <bere...@gmail.com>
wrote ----
> My first guess would be no, it wouldn't hurt performance.
> Although I have limited experience on Fuseki (just using it as a second
> test endpoint to verify SPARQL compatibility),
> I am using a similar approach on a different RDF store (currently at about
> 650,000 graphs on one deployment)
> I would imagine that Graphs are indexed the same as S, P, or O on Fuseki.
>
> One of my main problems with this approach has also been the lack of UI
> support for administration of large number of graphs as you pointed out.
> This is something not specific to Fuseki, apparently there isn't enough
> demand for this use case.
>
> On Tue, Jan 17, 2017 at 9:59 AM, Conal Tuohy <conal.tu...@gmail.com> wrote:
>
> > I am working with Fuseki 2 using the SPARQL Named Graph protocol, and I
> > wondered if there are practical limits on the number of named graphs in the
> > graph store?
> >
> > I know that many people use Jena only with a very small number of distinct
> > graphs, and I noticed that Fuseki's own user interface really only works
> > well when the number of named graphs is small (less than a thousand). That
> > is not a big problem in itself, since I don't need to use that UI, but I'm
> > more concerned about performance or other limitations when the number of
> > graphs is much higher; on the order of a million graphs, or a few million.
> >
> > Can anyone reassure me? Has anyone had problems with large number of named
> > graphs, and if so, were you able to fix them?
> >
> > Thanks!
> >
> > Conal
> >
> > --
> > Conal Tuohy
> > http://conaltuohy.com/
> > @conal_tuohy
> > +61-466-324297
> >
>
