From: Andres Freund <andres@anarazel.de>
To: pgsql-performance@postgresql.org
Subject: Re: Building multiple indexes concurrently
Date: Fri, 19 Mar 2010 18:38:08 +0100
User-Agent: KMail/1.13.1 (Linux/2.6.34-rc1-andres-01012-g1097d7c; KDE/4.4.1;
	x86_64; ; )
Cc: Hannu Krosing <hannu@2ndquadrant.com>, Greg Smith <greg@2ndquadrant.com>,
	Alvaro Herrera <alvherre@commandprompt.com>,
	Rob Wultsch <wultsch@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>
References: <2c5ef4e31003161804l5e32ceb9v6b7019118f30004d@mail.gmail.com>
	<4BA1403D.9030402@2ndquadrant.com>
	<1268860727.19220.1089.camel@hvost>
In-Reply-To: <1268860727.19220.1089.camel@hvost>
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Message-Id: <201003191838.09087.andres@anarazel.de>

On Wednesday 17 March 2010 22:18:47 Hannu Krosing wrote:
> On Wed, 2010-03-17 at 16:49 -0400, Greg Smith wrote:
> > Alvaro Herrera wrote:
> > > Andres Freund escribi=C3=B3:
> > >> I find it way much easier to believe such issues exist on a tables in
> > >> constrast to indexes. The likelihood to get sequential accesses on an
> > >> index is small enough on a big table to make it unlikely to matter
> > >> much.
> > >=20
> > > Vacuum walks indexes sequentially, for one.
> >=20
> > That and index-based range scans were the main two use-cases I was
> > concerned would be degraded by interleaving index builds, compared with
> > doing them in succession.
>=20
> I guess that tweaking file systems to allocate in bigger chunks help
> here ? I know that xfs can be tuned in that regard, but how about other
> common file systems like ext3 ?
ext4 should do that now by allocating the space for the files only after so=
me=20
time or uppon things like fsync (xfs does the same).
ext3 has, as far as I know, neither the ability to change allocation size n=
or=20
can do delayed allocation.

Andres