Date: Thu, 02 Mar 2006 11:34:40 +0700 From: Beastie <beastie@mra.co.id> To: Alex Zbyslaw <xfb52@dial.pipex.com> Cc: Liste FreeBSD <freebsd-questions@freebsd.org> Subject: Re: SATA Raid (stress test..) Message-ID: <440675E0.1020204@mra.co.id> In-Reply-To: <44058D9E.3010801@dial.pipex.com> References: <61560.207.70.139.52.1139628926.squirrel@www.compedgeracing.com> <6C543300-579C-4313-A355-0F26181B3FA7@todoo.biz> <067101c63216$a7682f10$c801a8c0@nexpc> <65260.207.70.139.52.1139998857.squirrel@www.compedgeracing.com> <06b901c63220$3a849eb0$c801a8c0@nexpc> <50778.207.70.139.52.1140002253.squirrel@www.compedgeracing.com> <43F3EDD6.80707@mra.co.id> <44052663.7000802@mra.co.id> <440565FF.3030002@mra.co.id> <44058D9E.3010801@dial.pipex.com>
next in thread | previous in thread | raw e-mail | index | archive | help
>
> Your performance sucks because, to quote the manual, "Input data is
> read and written in 512-byte blocks".
>
> Try a sensible blocksize. 16k would mimic a standard file system
> block, but even that is likely to underestimate. If you were, say,
> copying the disk to another you could easily use 1Mb.
>
> Some examples:
>
> dd if=/dev/ad10s1a of=/dev/null
> ^C794830+0 records in
> 794830+0 records out
> 406952960 bytes transferred in 164.049297 secs (2480675 bytes/sec)
>
> dd if=/dev/ad10s1a of=/dev/null bs=16k
> ^C53745+0 records in
> 53745+0 records out
> 880558080 bytes transferred in 21.092098 secs (41748245 bytes/sec)
>
> So from 2Mb/s to 41Mb/s!
>
> dd if=/dev/ad10s1a of=/dev/null bs=1m
> ^C933+0 records in
> 933+0 records out
> 978321408 bytes transferred in 13.836165 secs (70707556 bytes/sec)
>
>
> And up to 70Mb/s though nothing real world is likely to achieve that.
>
>
> There are a whole slew of ports (/usr/ports/benchmarks) some of which
> do disk tests. I've used unixbench in the past, which is a bit of a
> faff and does more than disks, but it works. If you run windows on
> the box and want graphical benchtests, then there are free apps out
> there that will do tests on disks, like Sandra.
>
> --Alex
>
second tools is diskinfo, but i'm not quite happy with the result.
#diskinfo -t /dev/amrd0s1d
/dev/amrd0s1d
512 # sectorsize
999996609024 # mediasize in bytes (931G)
1953118377 # mediasize in sectors
121575 # Cylinders according to firmware.
255 # Heads according to firmware.
63 # Sectors according to firmware.
Seek times:
Full stroke: 250 iter in 5.233346 sec = 20.933 msec
Half stroke: 250 iter in 3.828152 sec = 15.313 msec
Quarter stroke: 500 iter in 6.232849 sec = 12.466 msec
Short forward: 400 iter in 2.409001 sec = 6.023 msec
Short backward: 400 iter in 2.594473 sec = 6.486 msec
Seq outer: 2048 iter in 0.638372 sec = 0.312 msec
Seq inner: 2048 iter in 0.671994 sec = 0.328 msec
Transfer rates:
outside: 102400 kbytes in 1.102065 sec = 92916 kbytes/sec
middle: 102400 kbytes in 1.209657 sec = 84652 kbytes/sec
inside: 102400 kbytes in 1.912485 sec = 53543 kbytes/sec
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?440675E0.1020204>