From: Markus Roeffen (markus.roeffen_at_iws.uni-stuttgart.de)
Date: Tue Apr 03 2007 - 17:09:12 CEST
Message-ID: <46126E18.7040000@iws.uni-stuttgart.de> Date: Tue, 03 Apr 2007 17:09:12 +0200 From: Markus Roeffen <markus.roeffen@iws.uni-stuttgart.de> Subject: Re: Antw: Re: [suse-sles-e] Re: catastrophic disk-performance on SLES10 with SATA drives
Hi Andreas,
I've got a new benchmark and some comments for you.
First the benchmark on the external RAID:
server:/home # time dd if=/dev/zero of=data.out bs=8192 count=1310720
1310720+0 records in
1310720+0 records out
10737418240 bytes (11 GB) copied, 295,103 seconds, 36,4 MB/s
real 4m55.123s
user 0m0.588s
sys 0m32.142s
The number of pdflush processes during that procedure does _not_ increase.
But doing this as only user on the server the performance isn't good either.
matilda matilda wrote:
>> The same thing happens on our external Oxygen RAID system without cciss.
>>
>
>
>> And the performance is even worse. :-(
>>
>
> Do you also have the phenomena that the count of pdflush thread
> raises?
> And that the load is raising?
>
>
No pdflush, but load is raising.
Usually the I/O wait values are too high.
>> It was RAID1 and I was the only user. With more than 5 users the server
>>
>> is simply unusable, even my 11 year old Pentium Pro (without X11 and
>> KDE) would beat the poor performance without difficulties.
>>
>
> You have to be fair. If you have massive parallel access on a
> SOHO-Disk
> also nothing is happening any more. ;-) Probably this disk starts to
> burn.
>
>
You are right that it isn't a good idea to use SOHO equipment, but
setting up my old Pentium Pro as a temporary test server for website
projects (visited by one user of course), I usually had better performance.
Actually the performance on the ProLiant reminds me of making a shared
folder on a cd drive under Windows95 and there are more than 3 users,
who want to see, what's on the cd.
>> A new firmware version which isn't on the support cd has arrived (
>> http://h18023.www1.hp.com/support/files/server/us/revision/8926.html),
>>
>> so if the guys from Novell or HP (or from my external RAID storage) have
>>
>> no idea, I'll probably try that version.
>>
>
> I would like to hear if this has a positiv (probably p stand s for
> that)
> effect.
>
>
Sure, I'll let you know.
>> We also could test another linux distribution (Knoppix CD or something
>>
>
>
>> like that) to exclude (or verify) a SLES-only problem.
>>
>
> I had this idea too, but Knoppix doesn't ship with a cciss driver when
> I
> tried it the last time. Probably you find it on the newest DVD.
> I don't have a test machine here to try ubuntu.
>
>
If anything else fails, we still could try to remaster a knoppix cd with
an optional HP-PSP-patched SLES kernel. But that's a too huge task (with
an uncertain end) for me at the moment and I doubt, it's worth the effort.
> Anyway, I would like to hear if you find a solution or if someone is
> saying that this is "normal".
>
>
Actually I want to hear such a clear statement from one of the support
teams (HP, Novell, Oxygen). I'll let you know, when I've new information
from the support.
Best regards
Markus
-- ------------------------------- Markus Roeffen Institut für Wasserbau IWS Universität Stuttgart Pfaffenwaldring 61 70569 Stuttgart Tel: +49(0)711/685-67010 Fax: +49(0)711/685-67020 Mail: markus.roeffen@iws.uni-stuttgart.de --------------------------------------------------------------------- To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com For additional commands, e-mail: suse-sles-e-help@suse.com
This archive was generated by hypermail 2.1.7 : Tue Apr 03 2007 - 19:13:04 CEST