Date: Sun, 10 Sep 2006 11:32:20 +1200 From: Mark Kirkwood <markir@paradise.net.nz> To: Steven Hartland <killing@multiplay.co.uk> Cc: amd64@freebsd.org, Jo Rhett <jrhett@svcolo.com>, stable@freebsd.org Subject: Re: suggestions for SATA RAID cards Message-ID: <45034F04.9010103@paradise.net.nz> In-Reply-To: <079e01c6d417$3890fc40$b3db87d4@multiplay.co.uk> References: <44EC0B9B.5020705@withagen.nl> <003f01c6c68d$64688e60$b3db87d4@multiplay.co.uk> <20060907184316.GC56998@svcolo.com> <035701c6d2c3$eb574aa0$b3db87d4@multiplay.co.uk> <FEDA1103-8D83-4D43-9731-7E3D9D2DB1E5@svcolo.com> <001001c6d327$25dc07c0$b3db87d4@multiplay.co.uk> <4502123D.705@paradise.net.nz> <079e01c6d417$3890fc40$b3db87d4@multiplay.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
Steven Hartland wrote:
> Mark Kirkwood wrote:
>>
>> If you are using RAID0|5, then something is slowing you down (possible
>> clash between disk firmware and the Areca, or unfortunate choice of
>> strip chunk size).
>
> Dont know which test I was remembering but just did a quicky:
> OS: FreeBSD 6.1
> RAID: 5 on 5 * 400GB Seagate
> Controller: HighPoint 1820a
> CPU: Dual Opteron 244
> RAM: 2Gb
> /usr/bin/time -h dd if=/dev/da0 of=/dev/null bs=1048576 count=10000
> 10000+0 records in
> 10000+0 records out
> 10485760000 bytes transferred in 44.887239 secs (233602250 bytes/sec)
> 44.88s real 0.03s user 2.40s sys
>
> In comparison:
> OS: FreeBSD 5.4
> RAID: 5 on 6 * 300GB Seagate
> Controller: Areca 1120
> CPU: Dual Opteron 248
> RAM: 4Gb
> /usr/bin/time -h dd if=/dev/da0 of=/dev/null bs=1048576 count=10000
> 10000+0 records in
> 10000+0 records out
> 10485760000 bytes transferred in 81.598938 secs (128503633 bytes/sec)
> 1m21.60s real 0.00s user 2.69s sys
>
Hmmm - I've found that FreeBSD 6.1 is a considerably better performer
than 5.4, so that is not helping the comparison above.
Another thing to check is that both systems have the same vfs.read_max
sysctl tunable, as that makes quite a difference on RAID systems!
If you have the time to keep playing with these machines, it might be
interesting to try block sizes other than 1M - could expose different
behavior too!
Cheers
Mark
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45034F04.9010103>