The out-of-the-box Jena reasoners are in-memory reasoners. Reasoning
over TDB just makes them slower, doesn't allow them to scale better.
Things to try are:
- use reasoner OWLMicro (if that covers sufficient cases for your
purposes)
- allocate more memory
- use Pellet
- handcrafted inferences using SPARQL Update or java if you only need
some specific simple inferences that you can handle that way
Dave
On 22/03/14 10:29, Jacek Grzebyta wrote:
Dear All,
I have problem with owl reasoning. Even quite simple sparql request makes
huge CPU consumption and finally produces: java.lang.OutOfMemoryError:
GCoverhead limit exceeded
As a schema I use Uniprot owl (ftp://ftp.uniprot.org/pub/databases/uniprot
/current_release/rdf/core.owl) file loaded into TDB. There was no error
during that file loading. My config file is available here:
config.ttl<https://docs.google.com/file/d/0B9vmnMuEyN3VajBJNTZ1UW03SWc/edit?usp=drive_web>
And log file example here:
error.log<https://docs.google.com/file/d/0B9vmnMuEyN3VNVIwY3VYM0V6NDg/edit?usp=drive_web>
Is any way to optimise configuration? I know I have a lot of files for the
last graph but it is necessary to not to load into TDB. Because some of the
files would be replaced by newer versions. Is my own Java program with the
reasoning would have better performance?
Thanks a lot,
Jacek
