Re: Apache Marmotta

[Sergio Fernández <wi...@apache.org>]

Hi Oraskari,

I CC the users mailing list. Please, contact us via the public mailing
lists for any aspect regarding the Apache project. You can find the
subscription details at http://marmotta.apache.org/mail-lists.html

On 25 August 2015 at 16:32, Oraskari Jyrki <jy...@aalto.fi> wrote:
> We have been using Apache Marmotta since last winter, and it is a very
> nice piece of software. I like the software a lot.

Thanks for such nice words.

> However, our main concern is
> to be able to import large datasets (one million triples) fast in the triple
> store. Currently, the loading times are not preferable. As an expert,
> would you have an idea of is it possible and how would one import so
> large datasets into a triple store like Marmotta just in some minutes
> (preferably in seconds)?

Probably the first thing to get proper performance is to swtich to a
proper database as Marmotta backend. If you are still using the
default H2, it is extremely recommend to use PostgreSQL in real
environments:

http://marmotta.apache.org/configuration#db

At the wiki you can find all methods available to import data in Marmotta:

http://wiki.apache.org/marmotta/ImportData

>From those described there, the KiWiLoader is the  most performance
one since it directly goes to the database, and the recommended one
for really large datasets (for instance Freebase, that contains 2
billion triples). Depending of your hardware setup (SSD disk makes a
huge difference, for example) you can get imported around 10k
triples/sec.

> https://github.com/apache/marmotta/graphs/contributors does not show > much activity lately. Can we expect to have a new version of Marmotta
> this year?

That's right, besides the new GeoSPARQL support, code base is quite
stable and we have not that much in terms of new features. We did not
discuss it yet at dev@, but probably we'll prepare a new release at
some time Autumn.

Cheers,