[Pellet-users] Pellet query performance

[Markus Stocker]

Hi all,

I'm (still) rather new to Pellet. I have been testing ABox SPARQL
querying with Pellet (essentially the LUBM queries [1]) and I believe,
there is a significant performance difference between the following
two use cases. First, I created an inference model of the LUBM ABox
(as a separate process) and I did store it back to disk. The new ABox
contains all the entailed statements. In the second use case, I load
both the TBox and the ABox and I create an inference model with the
Pellet reasoner. Then I execute the queries over the inference model.

The queries are much faster in the second use case. I believe because
Pellet creates the required statistics to optimize the queries during
the reasoning process. In the first use case, reasoning is performed
as a separate process. Therefore, Pellet lacks the statistics.
However, result set sizes of the queries are correct in both use
cases. Is this statement correct?

Thanks,
markus

[1] http://swat.cse.lehigh.edu/projects/lubm/