Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 28 Aug 2010 19:50:24 +0400
From:      Ilya Bakulin <webmaster@kibab.com>
Cc:        Dag-Erling =?utf-8?Q?Sm=C3=B8rgrav?= <des@des.no>, hackers@freebsd.org, geom@freebsd.org
Subject:   Re: Support for WD Advanced Format disks
Message-ID:  <20100828195024.3d671a76@kibab.com>
In-Reply-To: <20100828130912.48205a47@kibab.com>
References:  <86wrry1hwv.fsf@ds4.des.no> <20100828130912.48205a47@kibab.com>

next in thread | previous in thread | raw e-mail | index | archive | help
--Sig_/1tkeDJKIRhJAOzIOcxqbVjm
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

On Sat, 28 Aug 2010 13:09:12 +0400
Ilya Bakulin <webmaster@kibab.com> wrote:

> Right now I'm running phybs -w /dev/ad7p1

Well, phybs finished.
Results:
kibab-desktop# ./phybs -w /dev/ad7p1
   count    size  offset    step        msec     tps    kBps

  131072    1024       0    4096     3098125       0      42
  131072    1024     512    4096     3100047       0      42

   65536    2048       0    8192     1505809       0      87
   65536    2048     512    8192     1505179       0      87
   65536    2048    1024    8192     1506127       0      87

   32768    4096       0   16384     1522343       1      86
   32768    4096     512   16384     1521017       1      86
   32768    4096    1024   16384     1523748       1      86
   32768    4096    2048   16384     1522853       1      86

   16384    8192       0   32768      890911       3     147
   16384    8192     512   32768      892026       3     146
   16384    8192    1024   32768      890699       3     147
   16384    8192    2048   32768      891862       3     146
   16384    8192    4096   32768      889903       3     147

Obviously 34-sectors offset is bad for phybs. Write requests are NEVER get =
aligned.
I have recreated partition as follows:

kibab-desktop# gpart delete -i 1 ad7
ad7p1 deleted
kibab-desktop# gpart add -t freebsd-ufs -s 10G -b 63 ad7
ad7p1 added
kibab-desktop# gpart show ad7
=3D>        34  2930277101  ad7  GPT  (1.4T)
          34          29       - free -  (15K)
          63    20971520    1  freebsd-ufs  (10G)
    20971583  2909305552       - free -  (1.4T)

So, we have ad7p1 which is still misaligned, but if write request will star=
t at offset=3D512, it will be aligned properly.
Let's test as follows:

kibab-desktop# ./phybs -l 4096 -w /dev/ad7p1
   count    size  offset    step        msec     tps    kBps

   32768    4096       0   16384     1520609       1      86
   32768    4096     512   16384       17898     151    7323
   32768    4096    1024   16384     1524899       1      85

:-) Aligned writes are 86 times faster than misaligned!!!
--=20
Regards,
Ilya Bakulin
http://kibab.com
xmpp://kibab612@jabber.ru

--Sig_/1tkeDJKIRhJAOzIOcxqbVjm
Content-Type: application/pgp-signature; name=signature.asc
Content-Disposition: attachment; filename=signature.asc

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.15 (FreeBSD)

iEYEARECAAYFAkx5MEUACgkQo9vlj1oadwhRAACgqRCXTcKsfesxaMxPjO/Hs9E4
iV8AmwQWE3Gs0EHsyJj6CAtoq4QgbP2L
=WxjB
-----END PGP SIGNATURE-----

--Sig_/1tkeDJKIRhJAOzIOcxqbVjm--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20100828195024.3d671a76>