Hi Bert!

There is no hard limit on the number of triples SwiftOWLIM can handle. 
However, this OWLIM flavor is meant to be fast and practical for not more 
than several millions of triples (certainly not 100M!), that is - it is fast 
on smaller datasets and doesn't scale very well beyond that. If your needs 
extend further than that, you don't need more RAM but BigOWLIM :)

On a sidenote, are you importing the Geonames dataset using some ruleset? Have 
in mind that inference might increase a lot the size of the dataset (e.g. you 
might be importing 200M instead of 100M triples actually).


Cheers,
Ivan

On Wednesday 16 December 2009 17:37:09 Bert Verslyppe wrote:
> Dear all,
>
>
> Is there a hard limit on the number of triples SwiftOWLIM can handle?
>
> When I tried to import the full GeoNames ontology (~100M triples), the
> import process stalled about halfway (on my laptop; I gave the SwiftOWLIM
> process 3 GB of RAM, which is the maximum I can do). If I understand the
> documentation well, adding more RAM should increase the amount of triples
> SwiftOWLIM is able to handle. However, Fig. 3 in the documentation shows
> that the 4cOpt12g benchmark, which is performed using 4 times more RAM,
> stalls at about the same number of triples. Is this a coincidence, or is
> 50M triples the maximum SwiftOWLIM can handle (e.g. as a consequence of
> data structure design)?
>
>
> Kind regards,
> Bert Verslyppe
> _______________________________________________
> OWLIM-discussion mailing list
> [email protected]
> http://ontotext.com/mailman/listinfo/owlim-discussion


_______________________________________________
OWLIM-discussion mailing list
[email protected]
http://ontotext.com/mailman/listinfo/owlim-discussion

Reply via email to