Message-ID: <28570856.post@talk.nabble.com>
Date: Sat, 15 May 2010 13:41:44 -0700 (PDT)
From: "joao.pinheiro" <joao.pinheiro@pontosi.pt>
To: pgsql-performance@postgresql.org
Subject: Re: Benchmark with FreeBSD 8.0 and pgbench
In-Reply-To: <4BEEFA75.6090708@agliodbs.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
References: <28569544.post@talk.nabble.com> <4BEEFA75.6090708@agliodbs.com>


Hi,
The tests were made without the -s parameter, (so 1 is assumed). I'm runnin=
g
the numbers
again on CentOS, with the optimized config and I'll test also different
scale values. I also will be able
to repeat the test again in FreeBSD with ZFS with the new options and
different scale,
but probably I won't repeat the UFS tests. I'll post the data when I finnis=
h
the tests.=20

Thanks for your feedback,
  Jo=C3=A3o Pinheiro


Josh Berkus wrote:
>=20
> Joao,
>=20
> Wow, thanks for doing this!
>=20
> In general, your tests seem to show that there isn't a substantial=20
> penalty for using ZFS as of version 8.0.
>=20
> If you have time for more tests, I'd like to ask you for a few more
> tweaks:
>=20
> (1) change the following settings according to conventional wisdom:
> =09wal_buffers =3D 8MB
> =09effective_cache_size =3D 9GB
> =09checkpoint_segments =3D 32
> =09on ZFS only: full_page_writes=3Doff
>=20
> (2) What scale were you using for the pgbench database?  I didn't see it=
=20
> in the e-mail.  It would be worth testing:
> =09s =3D 10 (small database, in memory)
> =09s =3D 500 (7GB, ram mostly full)
> =09s =3D 1000 (14GB, slightly larger than ram)
> =09s =3D 3000 (43GB, much larger than ram)
>=20
> If you were only testing a small size in your runs, then the only=20
> Filesystem behavoir you were testing was the transaction log.
>=20
> (3) Try a ZFS 128K record size
>=20
> (4) Centos/Ext3 appears to have had better staying power with high=20
> numbers of clients.  Can you continue testing with 50, 100 and 200=20
> clients on that combination?  And with data=3Dwriteback,noatime on Ext3?
>=20
> --=20
>                                    -- Josh Berkus
>                                       PostgreSQL Experts Inc.
>                                       http://www.pgexperts.com
>=20
> --=20
> Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org=
)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-performance
>=20
>=20

--=20
View this message in context: http://old.nabble.com/Benchmark-with-FreeBSD-=
8.0-and-pgbench-tp28569544p28570856.html
Sent from the PostgreSQL - performance mailing list archive at Nabble.com.