incubator-jena-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Vesse <>
Subject Re: DB2, RDF, and Jena?
Date Tue, 03 Apr 2012 23:38:20 GMT
That last comment applies both to comparisons between uRiKA and DB2 and between TDB and DB2


On Apr 3, 2012, at 3:11 PM, Robert Vesse wrote:

I think the Cray uRiKA comment seems to be a random aside

We don't run TDB anyway, we have a completely custom database and query engine.  Fuseki is
used purely as a way to wrap the database as a SPARQL endpoint.

I would want to see IBM's numbers for LUBM and SP2B before trying to make any real comparisons

Rob Vesse --<> -- A Division of Cray Inc
Software Engineer, Bay Area
m: 925.960.3941  |  o: 925.264.4729 | @:<>
 |  Skype: rvesse
6210 Stoneridge Mall Rd  |  Suite 120  | Pleasanton CA, 94588

On Apr 3, 2012, at 2:59 PM, Marco Neumann wrote:

So Timothy Prickett Morgan claims that DB2 outperforms TDB in a Cray
Urika setup by factor 3.5 on some unspecified SPARQL 1.0 query?

On Tue, Apr 3, 2012 at 5:48 PM, Andy Seaborne <<>>

The Apache Jena project is a Java framework for building semantic web
applications based on graph data, and Apache Fuseki is the SPARQL server
that processes the SPARQL queries and spits out the relationships so they
can be visualized in some fashion. (Cray's new Urika system, announced in
March, runs this Apache graph analysis stack on top of a massively
multithreaded server.)

Just like they imported objects and XML into the DB2 database so they could
be indexed and processed natively, IBM is now bringing in the RDF format so
that graph triples can be stored natively.

As IBM explains it – not strictly grammatically, to some English majors – a
triple has a noun, a verb, and a predicate, such as Tim (noun) has won
(verb) the MegaMillions lottery (predicate). You can then query all aspects
of a set of triples to see who else has won MegaMillions – a short list, in
this case.

In tests among DB2 10.1 early adopters, applications that used these graph
triples ran about 3.5 times faster on DB2 than on the Jena TDB data store
(short for triple database, presumably) with SPARQL 1.0 hitting it for

(Dear Reg, TDB does not stand for "triple database")



Marco Neumann

Join us at SemTech Biz in San Francisco June 3-7 2012 and save 15%
with the lotico community discount code 'STMN'

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message