On Wed, 2005-08-17 at 08:36 +0000, laurenz_brein wrote:

> However, when I run only these 'simple selects', the performance
> is a mere 50% worse than that of the worst competitor.
> I guess that the extra load the COUNT(*)-statements imply will
> explain the factor 4, and the 50% may be accounted for by the
> record versioning overhead, right?

Actually, this is accounted for by the 20 years of various companies'
ownership and varying design philosophies, before opening Firebird to
the open source community.

The Vulcan rewrite strips out much of the 20 years accumulation of
garbage for an improvement of 3 to 4 times in performance. Download Jim
Starkey's documents on Vulcan for a full technical and historical
explanation. I believe that they may be found on the IBPhoenix site.

Also, check out a Firebird 2 beta. It is reported to be much faster
because of a new on-disk structure.