From: Jon Schewe <jpschewe@mtu.net>
To: Andres Freund <andres@anarazel.de>
Cc: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-performance@postgresql.org
Subject: Re: How filesystems matter with PostgreSQL
Date: Fri, 04 Jun 2010 13:26:27 -0500
Message-ID: <4C094553.7040000@mtu.net> (raw)
In-Reply-To: <201006041633.23343.andres@anarazel.de>
References: <4C08EEDF.7090403@mtu.net>
<201006041504.03218.andres@anarazel.de>
<2503.1275661530@sss.pgh.pa.us>
<201006041633.23343.andres@anarazel.de>
On 6/4/10 9:33 AM, Andres Freund wrote:
> On Friday 04 June 2010 16:25:30 Tom Lane wrote:
>
>> Andres Freund <andres@anarazel.de> writes:
>>
>>> On Friday 04 June 2010 14:17:35 Jon Schewe wrote:
>>>
>>>> XFS (logbufs=8): ~4 hours to finish
>>>> ext4: ~1 hour 50 minutes to finish
>>>> ext3: 15 minutes to finish
>>>> ext3 on LVM: 15 minutes to finish
>>>>
>>> My guess is that some of the difference comes from barrier differences.
>>> ext4 uses barriers by default, ext3 does not.
>>>
>> Or, to put it more clearly: the reason ext3 is fast is that it's unsafe.
>>
> Jon: To verify you can enable it via the barrier=1 option during mounting..
>
>
>
First some details:
Linux kernel 2.6.31
postgres version: 8.4.2
More test results:
reiserfs: ~1 hour 50 minutes
ext3 barrier=1: ~15 minutes
ext4 nobarrier: ~15 minutes
jfs: ~15 minutes
--
Jon Schewe | http://mtu.net/~jpschewe
If you see an attachment named signature.asc, this is my digital
signature. See http://www.gnupg.org for more information.
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-performance@postgresql.org
Cc: jpschewe@mtu.net, andres@anarazel.de, tgl@sss.pgh.pa.us
Subject: Re: How filesystems matter with PostgreSQL
In-Reply-To: <4C094553.7040000@mtu.net>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox