Date: Thu, 15 Nov 2012 10:49:53 -0500
From: Bruce Momjian <bruce@momjian.us>
To: Craig Ringer <craig@2ndQuadrant.com>
Cc: Denis <socsam@gmail.com>, pgsql-performance@postgresql.org
Subject: Re: Thousands databases or schemas
Message-ID: <20121115154953.GA13919@momjian.us>
References: <1352367376704-5731189.post@n5.nabble.com>
	<CAEV0TzAv=BLZODAARJRwOAjQ3B=MuoZ+WntKfa4zO0OQ0wReCw@mail.gmail.com>
	<1352381380303-5731203.post@n5.nabble.com>
	<509C9F91.4020902@2ndQuadrant.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <509C9F91.4020902@2ndQuadrant.com>
User-Agent: Mutt/1.5.20 (2009-06-14)

On Fri, Nov  9, 2012 at 02:15:45PM +0800, Craig Ringer wrote:
> On 11/08/2012 09:29 PM, Denis wrote:
> > Ok guys, it was not my intention to hurt anyone's feelings by mentioning
> > MySQL. Sorry about that.
> It's pretty silly to be upset by someone mentioning another DB product.
> I wouldn't worry.
> > There simply was a project with a similar
> > architecture built using MySQL. When we started the current project, I have
> > made a decision to give PostgreSQL a try.
> It's certainly interesting that MySQL currently scales to much larger
> table counts better than PostgreSQL appears to.
> 
> I'd like to see if this can be improved down the track. Various people
> are doing work on PostgreSQL scaling and performance, so with luck huge
> table counts will come into play there. If nothing else, supporting
> large table counts is important when dealing with very large amounts of
> data in partitioned tables.
> 
> I think I saw mention of better performance with higher table counts in
> 9.3 in -hackers, too.

Yes, 9.3 does much better dumping/restoring databases with a large
number of tables.  I was testing this as part of pg_upgrade performance
improvements for large tables.  We have a few other things we might try
to improve for 9.3 related to caching, but that might not help in this
case.

-- 
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + It's impossible for everything to be true. +