RE: Performance issues with OWL Reasoners

> Not at doing DL reasoning. Relational databases do relational stuff
> well. For everything else, they are as likely to be rubbish as fast.

[VK] Agreed! But the hypothesis is that mapping into a proven scalable
technology such as an RDBMS, even if as a component helps build a scalable
DL reasoner. In general there is a "software development design pattern" that is
emerging which seems to be popular and may be working well.

1. XML Databases are being implemented by mapping XML querying operations into
an RDBMS and SQL queries
2. RDF Databases are being implemented by mapping RDF querying operations into
an RDBMS and SQL queries
3. In the past rule engines such as LDL++ have mapped into RDBMS technologies
for fact reasoning
4. Past examples of DLs (admittedly poorer versions than OWL-DL) being mapped
into RDBMS for ABox reasoning (CLASSIC, OBSERVER and others)

This is what is exciting about Instance store that it uses RDBMS and SQL queries
as a subcomponent (generating candidate answers) and then does post-processing
after that.

In all the above examples, each of them had to obviously implement something
extra, but RDBMS+SQL was leveraged as an important component.

I think the SW community should seek to leverage known scalable technologies to
reach industrial strength scalability and performance.

Cheers,

---Vipul

Received on Friday, 15 September 2006 14:12:34 UTC