novnov <novnovice@gmail.com> writes:
> http://www.simple-talk.com/sql/sql-server-2005/sql-server,-postgressql-and-fish-curry/
> at the bottom of the page is a post by Andrew Clarke that says pgsql is much
> slower in comparison to ms sql.
With no details, it's hard to rate that as anything but content-free FUD.
I don't doubt that he saw a serious performance differential on whatever
his test case was, but without seeing the test case it's impossible to
know why, or whether it would be fixable. Nor do we know what PG
version he was testing. Given that he mentions Windows 2000, it's not
unlikely that he was testing a Cygwin port of PG 7.something, in which
case just moving to an 8.1 native port would help a lot.
Also, given that he mentions that complex joins were bad, it could be
that he forgot to analyze, or had a corner case where the row estimates
were bad anyway, or had an outer-join scenario where reordering of
the outer joins is really essential for performance. (We've fixed the
latter as of 8.2, I think, but I will agree it's a serious weak spot in
existing PG releases.) But this is all speculation...
regards, tom lane
