[Ocfs-devel] Re:Re: Re:Re:Re:Re:Re: [Ocfs-users] Re: Mode context extremely poor performance and varyio

[Michael Aubertin]

Stephen, all,

Thank u very much for your answer and i wish you an happy thanksgiving. 

Currently we tried to migrate our base on RAW devices already using e41smp, SecurePath (HP) 3C and Qlogic 7.00. The average response time of our sql request on ia32 is 0.32Sec very regularly. On our Itanium on OCFS that's very randomly between 1 and 15Sec, and on RAW between <1s and 3 sec. David is an another customer of RedHat,HP,Oracle and happen a bug very close to our. Then both HP and RedHat provide him a special kernel e47 and special SecurePath with the last 7.01 qlogic driver. That's the reason why i ask you for it. Because our name is provide by the top changed changelog in dot spec.

Thank you for your help.

Athos10.

PS: If you have an idea, it will be very welcome. 



Original message:
---------------
From:"Stephen C. Tweedie" <sct at redhat.com>
To:Michael Aubertin <maubertin at ares.fr>
Cc:Jason Baron <jbaron at redhat.com>, s_wendy_cheng at hotmail.com, David McWhinnie <davidmcwhinnie at yahoo.com>, jean-yves.bocahu at hp.com, igaste at ares.fr, a.brunel at ilius.net, ocfs-users at oss.oracle.com, ocfs-devel at oss.oracle.com, Stephen Tweedie <sct at redhat.com>

Hi,
On Wed, 2004-11-24 at 22:08, Michael Aubertin wrote:
> Hello, i write to you to having your opinion and may be a little help.
> It seem you write a patch on varyio stack in errata special kernel for
> Itanium. We currently have an issue using OCFS, Secure Path and Qlogic
> driver. According to the source i read, varyio is very close to this
> two driver especialy qlogic. Do you believe, like i begin to do
> regarding the source, that our performance issue can be solve by your
> patch. 
varyIO is generally there only to reduce CPU utilisation for raw IO, not
to improve actual disk throughput.  VaryIO does have a couple of other
roles --- it enables sector-granularity on ext3 O_DIRECT on later
kernels, for example.  But it was never intended or expected to result
in disk IO performance improvements.
If you just want to look at source code, then varyIO has been in our
mainline Red Hat Enterprise Linux kernels since AS-2.1 --- you can
download the src.rpms for AS-2.1 and RHEL3 from ftp.redhat.com.  Those
contain broken-out patches for various functionality.
For any information about special kernels outside the normal public
errata stream, you'll have to follow the chain of contacts upstream
which provided that kernel.
Cheers,
 Stephen




****************************************************************************************************
Ce message ou ses pieces jointes peuvent contenir des informations confidentielles a l'intention exclusive de son destinataire et est couvert par le secret professionnel.
Toute utilisation, divulgation ou reproduction de son contenu sont strictement interdits.
Si vous avez recu ce message par erreur, merci de le notifier a son expediteur et d'en detruire toute copie.
Le present message pouvant-etre altere a notre insu, le groupe ARES ne peut pas etre engage par son contenu.
www.ares.fr
****************************************************************************************************