Thread: XFS vs Ext3, and schedulers, for WAL

XFS vs Ext3, and schedulers, for WAL

From
Josh Berkus
Date:
All,

There's a number of blog tests floating around comparing XFS and Ext3,
and the various Linux schedulers, for PGDATA or for an all-in-one mount.

However, the WAL has a rather particular write pattern, and it's
reasonable to assume that it shouldn't be optimized the same way as
PGDATA.  Has anyone done any head-to-heads for WAL drive configuration
changes?

--
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com

Re: XFS vs Ext3, and schedulers, for WAL

From
Josh Berkus
Date:
> There's a number of blog tests floating around comparing XFS and Ext3,
> and the various Linux schedulers, for PGDATA or for an all-in-one mount.
>
> However, the WAL has a rather particular write pattern, and it's
> reasonable to assume that it shouldn't be optimized the same way as
> PGDATA.  Has anyone done any head-to-heads for WAL drive configuration
> changes?

That would be a "no", then.  Looks like I have my work cut out for me ...

--
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com

Re: XFS vs Ext3, and schedulers, for WAL

From
"Joshua D. Drake"
Date:
On Mon, 2010-10-11 at 10:50 -0700, Josh Berkus wrote:
> > There's a number of blog tests floating around comparing XFS and Ext3,
> > and the various Linux schedulers, for PGDATA or for an all-in-one mount.
> >
> > However, the WAL has a rather particular write pattern, and it's
> > reasonable to assume that it shouldn't be optimized the same way as
> > PGDATA.  Has anyone done any head-to-heads for WAL drive configuration
> > changes?
>
> That would be a "no", then.  Looks like I have my work cut out for me ...

The only thing I have done is:


http://www.commandprompt.com/blogs/joshua_drake/2008/04/is_that_performance_i_smell_ext2_vs_ext3_on_50_spindles_testing_for_postgresql/

It doesn't cover XFS but it provides a decent and simple comparison on
ext2/ext3 etc...

Remember xlog is sequential so pushing it off is useful.

JD

>
> --
>                                   -- Josh Berkus
>                                      PostgreSQL Experts Inc.
>                                      http://www.pgexperts.com
>

--
PostgreSQL.org Major Contributor
Command Prompt, Inc: http://www.commandprompt.com/ - 509.416.6579
Consulting, Training, Support, Custom Development, Engineering
http://twitter.com/cmdpromptinc | http://identi.ca/commandprompt


Re: XFS vs Ext3, and schedulers, for WAL

From
Josh Berkus
Date:
>
http://www.commandprompt.com/blogs/joshua_drake/2008/04/is_that_performance_i_smell_ext2_vs_ext3_on_50_spindles_testing_for_postgresql/
>
> It doesn't cover XFS but it provides a decent and simple comparison on
> ext2/ext3 etc...

Yeah, it doesn't test actual log writing, though.  Nor specific settings
even for those two filesystems.  So it's still a mystery ...

--
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com