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?
Always nice to be compared with.
TDB can improve by 3.5 on datasets in the size range it is more often
used for.
I'm having trouble finding the time to work solidly on it but I hope to
produce an experimental version using merge joins. More compact indexes
would be good as well - that's not for improved performance but for more
triples-per-gigabyte.
The other area to look at is simply to profile to find hotspots (in
sorting?) - they may be places where the current code is just plain
inefficient.
Andy
