AMI megaraid performance (was: RE: Dell release of Redhat 7.2)

Seth Mos knuffie at
Fri Nov 9 17:52:00 CST 2001

At 20:26 9-11-2001 +0000, Arjan van de Ven wrote:
> > > The 2.4.9-12 kernel is the RedHat kernel that was released as
> > > a security
> > > update.
> > > The 2.4.14 kernel is the linus kernel. The filesystem is XFS
> > > on both versions.

Oops, not intended result. Sorry Arjan.

>That is NOT a Red Hat kernel, if you are using XFS. It's an SGI XFS

I know. I maintain the XFS FAQ after all :-)

>which is versioned by SGI to resemble the version of the Red Hat kernel.
>it has a different VM and has several other things changed;

It has the -ac VM with XFS. It also needs some mainline kernel changes to 
go with that.
I explicitely mentioned XFS as the used fs to make that distictintion.

>  it's
>therefore incorrect
>to use SGI XFS's kernel for performance measurement and then attributing
>that to the
>real Red Hat Linux kernel.

I was trying to point out huge differences between the linus and -ac 
kernels in regard to the AMI megaraid performance. This has to do with 
relative speed improvements not with absolute.
You get a nice performance boost with the new firmware under ext2 as well. 
But since none of my servers use ext2 I can't test that.

Note that after some testing the 2.4.9-12 (-ac if you want) has a better IO 
throughput on multiple processes. The read speed is less but the whole is 
more balanced resulting in a faster dump and restore of our progress 8/9 
databases. This workset is about 10GB of database and the differnce between 
about 30 minutes vs. 45-60 minutes is kind of significant if you ask me.

>I know SGI often tries to project the appearance of being the Red Hat
>Linux kernel

No they don't.

>so that people think they benefit from Red Hat's testing and bugfixes;
>however this is a false

SGI test their own kernels before releasing them.

>use of the Red Hat brand. (The changes made by SGI to the Red Hat Linux
>kernel codebase,
>such as using known broken compilers, basically invalidate all testing).

What's broken about 2.91.66? SGI does their own testing of release kernels 
to ensure that stuf works. They are not trying to piggy back. They are 
making a redhat compatible installer and make the kernels look and feel as 
much like the official redhat kernels to prevent issues for people 
trying/using them.

Seth Mos
XFS FAQ maintainer
Every program has two purposes one for which
it was written and another for which it wasn't
I use the last kind.

More information about the Linux-PowerEdge mailing list