Dave Page <dpage@postgresql.org> writes:
> Josh Berkus wrote:
>> Higher performance (+20% in OLTP benchmarks)
> Do we actually know what Tweakers are doing? Their graphs and articles
> seem very good n'all, but the best I've been able to ascertain is that
> it's some in-house benchmark based on their own (99% read only?)
> website. Not that that's a bad thing, just not exactly OLTP...
Another problem is that the Tweakers result Josh is quoting is based on
"8.2devel", which turns out to mean a CVS pull from 3-Jun-2006, which is
quite a bit different from 8.2 final --- and Arjen says 8.2 final looks
a lot worse:
http://archives.postgresql.org/pgsql-performance/2006-12/msg00041.php
I've been discussing this with him off-list and trying to figure out
why such a drop, but no answers yet. I'm pretty hesitant to wave that
Tweakers article around until we know what's going on here. If we have
other benchmarks we can point to that justify the "20%" number, then
it's OK to use ...
regards, tom lane