[Linux-cluster] Re: 2.6.11-based GFS's concurrent-read performance declined badly

[Ken Preslan]

On Sat, Mar 19, 2005 at 09:21:58PM +0800, ?????? wrote:
> >Do you see a similar drop in performance when you run your test on
> >ext3 on the two kernels?      
>
> Yes, I've formated my gfs to ext3fs, the result is the same.
> And I've formated one local raw disk to ext3fs, 
> the 2.6.11-based ext3fs's performance declined too.
> So maybe it dues to my compilation of the 2.6.11.3 kernel?
> Which parameters will be so important that cause this problem?

I would be suprised if it was how you compiled the kernel or a change
in a user-tuneable parameter that caused the problem.  It was probably
something that changed in your storage driver or the core I/O layers
between the two kernel versions.

I'm not sure what's the best way of getting the problem fixed.  Complaining
on the linux-kernel list might help.

-- 
Ken Preslan <kpreslan at redhat.com>