We have 28m triples in an OWLIM-Lite (v5.3) store and have found SPARQL deletions very slow (15 minutes to delete 7 triples). We have tried turning off inference, as per the recommendation in: http://answers.ontotext.com/questions/734/archive-performant-deletion-query .

Is there anything else we can do to improve performance? We are currently using the reification vocabulary (rdf:Statement, rdf:subject etc) as we need to store additional pieces of information about the statements - can this cause performance problems? Or is OWLIM simply not well suited to frequent deletions?

Thanks for your help,

Cathy

asked 30 May '13, 10:03

Cathy's gravatar image

Cathy
16112
accept rate: 0%


I think you have guessed the answer. When using inference, OWLIM-Lite does NOT do incremental retraction, so any deletion requires a complete recomputation of the inferred closure.

Without inference, then this recomputation is not required and so should be faster. Just to check, how are you turning off inference? It might be necessary to create a new repository with the new configuration and reload it.

Regarding the reification vocabulary, this should not cause a problem, although you might also try using named graphs (context) instead.

link

answered 07 Jun '13, 09:40

baz's gravatar image

baz ♦♦
33128
accept rate: 36%

Thanks for replying Barry - however, what we found was:

1) We turned off inference, and reloaded all the data in the newly configured repository

2) We could then delete, but no longer query anything (no results were being returned at all.)

3) We wondered whether this was because OWLIM needs some inference to link the rdf:Statement triple to the rdf:subject triple in our reification structures - this doesn't quite make sense to me, but the results supported this - when we changed the schema to use OWL based reification (i.e. made our predicates into OWL classes themselves) and ditched rdf:Statement, rdf:subject etc, we could then both query and delete the triple store.

So unfortunately, it does seem like there's a problem with the reification vocabulary in OWLIM. Regarding using named graphs as an alternative, what is the performance like for querying when you have say 10 million named graphs? We'd need one per original triple, which seems a lot.

(07 Jun '13, 09:52) Cathy
Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Tags:

×37
×11
×2
×1

Asked: 30 May '13, 10:03

Seen: 5,003 times

Last updated: 07 Jun '13, 09:52

powered by BitNami OSQA