Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 28 Aug 2010 13:05:49 -0300
From:      Thiago Damas <tdamas@gmail.com>
To:        Ilya Bakulin <webmaster@kibab.com>
Cc:        =?ISO-8859-1?Q?Dag=2DErling_Sm=F8rgrav?= <des@des.no>, hackers@freebsd.org, geom@freebsd.org
Subject:   Re: Support for WD Advanced Format disks
Message-ID:  <AANLkTikwgiw7r1KFdMVYaoomMGmvpjw=2rGtqM%2BZ3at2@mail.gmail.com>
In-Reply-To: <20100828195024.3d671a76@kibab.com>
References:  <86wrry1hwv.fsf@ds4.des.no> <20100828130912.48205a47@kibab.com> <20100828195024.3d671a76@kibab.com>

next in thread | previous in thread | raw e-mail | index | archive | help

You read:
"ATA 4K sector issues"
http://lists.freebsd.org/pipermail/freebsd-hackers/2010-March/031154.html

2010/8/28 Ilya Bakulin <webmaster@kibab.com>:
> 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
> =>        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 start at offset=512, 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!!!
> --
> Regards,
> Ilya Bakulin
> http://kibab.com
> xmpp://kibab612@jabber.ru
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTikwgiw7r1KFdMVYaoomMGmvpjw=2rGtqM%2BZ3at2>