Re: [ANN] 4store version 1.0.0

Apologies for messing up threading, this mail didn't get through to me  
for some reason, I found it in the archives by chance.
> > Garlik is pleased to announce the release of 4store version 1.0.0.
> >
> > For more information see http://4store.org/
>
> Congratulations!   It looks like a really nice bit of work.
>
> I haven't really followed the state of the art in quad-store
> benchmarking.  The ESW page isn't totally stale, but it's not that  
> fresh
> either [1].  Do you have any figures that are worth sharing?
>
> I'm particularly interested in small & fast queries, for little web
> apps.  My intuition is that something based on SQLite would be faster
> for the small stuff, since there's no need for IPC.  But, of course,
> intuition on such things isn't good for much more than coming up with
> good questions to ask.
Our experience is that this approach of running SPARQL queries over  
HTTP is easily fast enough to support that kind of query.

For example the page at http://foaf.qdos.com/find/?q=timbl%40w3.org is  
built entirely from SPARQL queries over a FOAF KB (without something  
like 10M FOAF files in it), and requires several dozen SPARQL  
queries*, some of which are very complicated, but it generates the  
HTML in about .4 of a second. I'd be happy to share the exact queries  
that are run in the creation of this page offlist, but they're pretty  
bulky. They have to do the IFP joining up for all the foaf:knows  
people, so you can imagine that there's quite a few, and they're  
extremely complex.

3rd parties are currently working on benchmarks of 4store, we have  
internal benchmarks, but those wouldn't really be meaningful to anyone  
else at the moment, though we might consider publishing them at some  
point in the future - they cover ground that's not really captured by  
existing benchmarks.

Anecdotally I would say that the query performance is at on-par with  
other stores, and import performance is somewhat higher, but I'm not  
up to speed with current developments. Per-query, and per-data  
performance in RDF stores seems to vary substantially in SPARQL  
systems. We look forward to seeing some public benchmark results.

* There would be fewer queries, but it's built on SPARQL 1.0, which  
has no aggregates.

- Steve
> [1] http://esw.w3.org/topic/RdfStoreBenchmarking
-- 
Steve Harris
Garlik Limited, 2 Sheen Road, Richmond, TW9 1AE, UK
+44(0)20 8973 2465  http://www.garlik.com/
Registered in England and Wales 535 7233 VAT # 849 0517 11
Registered office: Thames House, Portsmouth Road, Esher, Surrey, KT10  
9AD

Received on Monday, 19 October 2009 11:08:34 UTC