Optimizer, partitions & attitudes.

I made a recent post on this mailing list which was largely ignored, not
many replies followed. However, I was accused of being acrimonious on
comp.databases.postgresql and it wasn't my intention to come across that
way. I humbly apologize to everybody whom I have offended and to all
those who I am still to offend. Sorry for the cross post, but I believe
that there is an attitude problem that has to be addressed. I am quoting
the Usenet post in its entirety:
****************************************************************************************************************************
On Thu, 02 Sep 2010 15:56:11 +0200, Anselmo Canfora wrote:

> BTW I have a 300 million rows
> partitioned table under PG and it performs really well, with just a
> little bit of attention in the way you issue queries. Surely you have
> the skills to get PG to do what you want, just do not pretend it to do
> the same things Oracle does. If you want advanced partitioning,
> replication, materialized views, flashback queries and so on, just put
> on table some ten thousand bucks (if will suffice) and you'll be happy
> and less acrimonious in the way you interact with other people. Come
> on... be nice.

Hmmm, I'm acrimonious? I've been called worse, but it wasn't my intention
to come across that way. I simply wanted to warn about something I
consider a serious bug. As for the comparison with Oracle, the biggest
problem that I see is performance, not the lack features. I am seriously
working on moving several TB-sized database to Postgres, but in order for
me to do that, Postgres would have to execute certain queries a lot
faster than it does. I am even willing to rewrite queries but if I am
ever to pass the UAT, the database must perform better than it currently
does.
Furthermore, I don't consider myself particularly caustic or acrimonious.
I do think that drawing attention to what looks like a bug is actually a
service to the open source community, not being a SOB.
After all, there was a very recent presentation about the optimizer which
looks as if it was written for little children and starts with "you
shouldn't assume that you're smarter than the optimizer". If the
presenter can insult his audience like this with a smile and tell them
that they shouldn't think that they're smarter than a computer program,
some among the audience may even fail the Turing test, why should I be
particularly tender and sensitive when it comes to database problems?
Joking aside, I am not a fan of "be nice, always smile" attitude. I am a
DBA, a guy who lives in the rough place, where the rubber meets the road.
I've always been rough around the edges, much more so with Oracle than
with Postgres. Postgres still has a long way to go, even to catch up with
Oracle7. Such sanctimonious attitude as you exhibit will not get things
done. In case you haven't noticed, Postgres is only gaining popularity
now and it is not yet a safe bet, not even in the world of OSS databases.
I fought for Postgres against the proponents of MySQL, an open source
database which is still more popular than Postgres. If Postgres wants to
overtake MySQL, let alone Oracle, some problems need to be solved. As an
end user, I will warn about the problems that I encounter. I consider
that a service to the community. As an immediate improvements, Postgres
would need the wait event interface, tracing facility and means of
influencing the optimizer.
Also, presenters of the optimizer related informations should consider
providing a really useful information and tweaking the parameters,
instead of insulting their audience by telling them that they shouldn't
assume that they're smarter than the optimizer.
**********************************************************************************************************************

--

Mladen Gogala
Sr. Oracle DBA
1500 Broadway
New York, NY 10036
(212) 329-5251
http://www.vmsinfo.com
The Leader in Integrated Media Intelligence Solutions