From: Andres Freund <andres@2ndquadrant.com>
Organization: 2ndquadrant.com
To: pgsql-performance@postgresql.org
Subject: Re: PostgreSQL db,
	30 tables with number of rows < 100 (not huge) - the fastest way to
	clean each non-empty table and reset unique identifier column
	of empty ones.
Date: Fri, 6 Jul 2012 16:14:26 +0200
User-Agent: KMail/1.13.7 (Linux/3.4.0-andres; KDE/4.8.4; x86_64; ; )
Cc: Daniel Farina <daniel@heroku.com>, Craig Ringer <ringerc@ringerc.id.au>,
	Stanislaw Pankevich <s.pankevich@gmail.com>
References: 
 <CAFXpGYbgmZYij4TgCbOF24-usoiDD0ASQeaVAkYtB7E2TYm8Wg@mail.gmail.com>
	<4FF6CC11.2000904@ringerc.id.au>
	<CAAZKuFbpYyfRA=+DKD4yCr_c_Tag2JNO32B8NVU15gg4xX3u_Q@mail.gmail.com>
In-Reply-To: 
 <CAAZKuFbpYyfRA=+DKD4yCr_c_Tag2JNO32B8NVU15gg4xX3u_Q@mail.gmail.com>
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Message-Id: <201207061614.26739.andres@2ndquadrant.com>

On Friday, July 06, 2012 01:38:56 PM Daniel Farina wrote:
> ll, I don't know a
> mechanism besides slow file system truncation time that would explain
> why DELETE would be significantly faster.
There is no filesystem truncation happening. The heap and the indexes get 
mapped into a new file. Otherwise rollback would be pretty hard to implement.

I guess the biggest cost in a bigger cluster is the dropping the buffers that 
were formerly mapped to that relation (DropRelFileNodeBuffers).

Andres
-- 
 Andres Freund	                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services