On Wed, Jun 27, 2007 at 08:01:12PM +0200, pierre-philipp braun wrote:
> (posted to tech-kern, not i386 & current-users)
> 
> Hi,
> 
> I've tryed NetBSD current (20070513 or 16) on a poweredge 2900, SAS Dell 
> PERC 5/i.  dmesg atteched.  It worked, but _very_ slowly,
> 
> bsdp# time dd if=/dev/zero of=lala.zero count=100 bs=1024k
> 100+0 records in
> 100+0 records out
> 104857600 bytes transferred in 7.251 secs (14461122 bytes/sec)
>         7.26 real         0.00 user         0.34 sys

What kind of disk and array do you have ? On dell 2950 I get about 70/80MB/s
on a raid5 array.
Do you have the battery backup for the cache ?

> [...]
> in comparaison, SLES9sp3 + Dell official drivers (I know, it's unfair),
> 
> slesp:~ # time dd if=/dev/zero of=lala.zero count=100 bs=1024k
> 100+0 records in
> 100+0 records out
> 
> real    0m0.310s
> user    0m0.000s
> sys     0m0.309s
> 
> 
> slesp:~ # sh bench.sh
> 100+0 records in
> 100+0 records out
> 
> real    0m0.503s
> user    0m0.000s
> sys     0m0.492s
> 100+0 records in
> 100+0 records out
> slesp:~ #
> real    0m0.734s
> user    0m0.000s
> sys     0m0.722s
> 
> 
> bench.sh:
> #!/bin/sh
> time dd if=/dev/zero of=lala.zero count=100 bs=1024k &
> time dd if=/dev/zero of=lala2.zero count=100 bs=1024k
> rm lala.zero
> rm lala2.zero

I suspect with this, you only check the memory bandwidth of the system.
The data may not reach the disk at all.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--