SAS 6/iR write performance depends on kernel version, blade server m600

LALOT Dominique dom.lalot at gmail.com
Wed Jun 4 10:34:13 CDT 2008


CONCLUSION!!


Well, it appears that in fact, there is a change in 2.6.22 about MM and
there is less memory for write caching. So if you stay under 80Mo, it works
well. After you get disk performance. We also check it against SATA devices
and we observed it was not related to hardware.

Sorry for such "noise". Why did I choose 600Mo!..

Dominique

2008/6/2 Dominique LALOT <dom.lalot at gmail.com>:

>  I have some more information about kernel version
>
> To be more precise: I/O writes are OK till 2.6.21. The problem appears in
> 2.6.22. I compiled more than 20 kernels since a few weeks now. Hopefully,
> the hardware is very fast except that I/O problem.
>
> debian-test:/usr/src# uname -a
> Linux debian-test.pr.univmed.fr 2.6.22.1 #1 SMP Mon Jun 2 12:03:44 CEST
> 2008 i686 GNU/Linux
> debian-test:/usr/src# dd if=/dev/zero of=/root/test.cdrom bs=10k
> count=60000
> 60000+0 enregistrements lus
> 60000+0 enregistrements écrits
> 614400000 octets (614 MB) copiés, 8,10601 seconde, 75,8 MB/s
> debian-test:/usr/src# time tar -xzf linux-2.6.24.tar.gz
>
> real    0m9.740s
> user    0m2.732s
> sys     0m1.520s
> debian-test:/usr/src# dmesg | grep -i write
> sd 4:1:0:0: [sdb] Write Protect is off
> sd 4:1:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't
> support DPO or FUA
> sd 4:1:0:0: [sdb] Write Protect is off
> sd 4:1:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't
> support DPO or FUA
>
>
> debian-test:~# uname -a
> Linux debian-test.pr.univmed.fr 2.6.21.7 #1 SMP Fri May 30 18:39:58 CEST
> 2008 i686 GNU/Linux
> debian-test:~# dd if=/dev/zero of=/root/test.cdrom bs=10k count=60000
> 60000+0 enregistrements lus
> 60000+0 enregistrements écrits
> 614400000 octets (614 MB) copiés, 1,21061 seconde, 508 MB/s
> debian-test:~#
> debian-test:/usr/src# time tar -xzf linux-2.6.24.tar.gz
>
> real    0m2.602s
> user    0m2.544s
> sys     0m1.508s
>
> debian-test:~# dmesg | grep -i write
> sda: Write Protect is off
> SCSI device sda: write cache: enabled, read cache: enabled, doesn't support
> DPO or FUA
> sda: Write Protect is off
> SCSI device sda: write cache: enabled, read cache: enabled, doesn't support
> DPO or FUA
> I'm not sure now that the problem comes from that string. The string has
> been changed/added recently
>
>
>
>
> Alessandro FAGLIA a écrit :
>
> -------- Original Message  --------
> Subject: Re: SAS 6/iR write performance depends on kernel version, blade
> server m600
> From: Dominique LALOT <dom.lalot at gmail.com> <dom.lalot at gmail.com>
> To: Alessandro FAGLIA <a.faglia at farmol.it> <a.faglia at farmol.it>, linux-poweredge at dell.com
> Date: Fri May 30 2008 10:34:05 GMT+0200 (ora legale Europa occidentale)
>
>
>  Alessandro FAGLIA a écrit :
>
>
>  -------- Original Message  --------
> Subject: SAS 6/iR write performance depends on kernel version, blade
> server    m600
> From: Dominique LALOT <dom.lalot at gmail.com> <dom.lalot at gmail.com>
> To: linux-poweredge at dell.com
> Date: Mon May 19 2008 16:05:10 GMT+0200 (ora legale Europa occidentale)
>
>
>  Hello,
>
> That's quite a long time, I'm "playing" around a performance problem.
>
> We just acquire some  new blades and  I discovered something strange
> affecting the new SAS 6 subsystem.
>
> [...]
>
>
>
>  Good morning Mr Lalot.
>
> I'm a Linux-PE ML reader, and we are buying two new R300 server with
> SAS 6iR RAID controller. We'd like to use them in RAID1 configuration.
>
> I read your thread named "SAS 6/iR write performance depends on
> kernel version, blade server m600". BTW, there are other threads in
> the lists say that Debian (any version) is not going to work on such
> controller.
>
>
>  We are all using linux, we should not say debian won't .. or redhat
> will.. What we like about debian is the packaging system and the very
> few packages installed at first which is perfect for servers.
> Debian Etch is working prefectly on LSI sas 6. That's a kernel issue.
> Fortunatly all the "Dell supported distrib" are using a kernel <2.6.18.
>
>  Now I understand. If I'm correct, we should simply compile a kernel
> version less than 2.6.18 and everything should be fine.
>
>
>
>  I'm still looking at kernel patch to see why sas 6  LSI controller is
> working so bad in the last kernels. I'm sure now, that the last
> kernels are asking firware, and the default settings are bad.
>
> What we can say, is that last Ubuntu server 8.04  behaves very badly
> with LSI sas 6 (2.6.24)
>
>
>  Thanks for your feedback.
>
> --
> Alessandro
>
>
> _______________________________________________
> Linux-PowerEdge mailing listLinux-PowerEdge at dell.comhttp://lists.us.dell.com/mailman/listinfo/linux-poweredge
> Please read the FAQ at http://lists.us.dell.com/faq
>
>
>


-- 
Dominique LALOT
Ingénieur Systèmes et Réseaux
http://annuaire.univmed.fr/showuser?uid=lalot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.us.dell.com/pipermail/linux-poweredge/attachments/20080604/1084d67e/attachment.htm 


More information about the Linux-PowerEdge mailing list