Knowing what people are thinking is important in improving the product. With that in mind I did a quick Google for people mentioning Kowari. I'll use it here to correct anything too.
DeliverableS4Simile "Naive use of persistent store in Jena decreases performance by 100x:
* Part of the problem is limited expressivity in RDQL
* Look at performance tuning on databases
* Ryan: Look at the performance of more specialized RDF databases, cf. Kowari"
Open Source Projects That Use Java NIO "The storage engine of Kowari is a transactional triplestore known as the XA Triplestore. ll relevant fields of in-memory and on-disk data structures are 64 bits wide..."
Kowari for hundreds of millions of triples "Jim Hendler emailed me in response to my having mentioned on firstname.lastname@example.org that I was surveying triple stores for use in data mining and machine learning. He mentioned a Java-based, non-relational, triple store called Kowari that is available in open source form..."
RDQLPlus "I discovered Kowari last week. The iTQL language is very similar to what I've come up with for RDQLPlus... having equally been inspired by SQL/DDL. Kowari looks like a nice database. I've downloaded it but haven't had a chance to play with it yet..."
Some Tools "Kowari is a layer on top of Jena with OWL reasoning, too"
I would say that Kowari is a layer beneath Jena - one that provides persistance. The OWL reasoning is really only offered at the moment through Jena. However, we will be getting some basic inferencing, at our own query layer, in there soon too.
[protege-discussion] Re: large data sets, bulk data acquisition "I had a really bad time with Kowari earlier this year, it wouldn't compile and then pass its own self-tests...."
This is basically problems with Windows. We develop on Linux and OS X and only do QA on Windows. Our initial release had known problems under Windows - which lead to failing unit tests but was not fatal for data storage. Anyway, it is fixed now; although Windows does have some drawbacks when it comes to using NIO.