On Thu, Apr 16, 2009 at 11:50:27AM +0200, Bart Van Assche wrote:
> On Wed, Apr 15, 2009 at 9:19 PM, Pasi Kärkkäinen <pa...@iki.fi> wrote:
> > noop is usually good for the initiator. cfq has a feature (or a bug?) that
> > prevents achieving queue depths deeper than 1, and thus limits your
> > bandwidth a lot when there are (or should be) many ios on the fly at the
> > same time.
> Do you remember on which kernel version you observed the above
> behavior ? This might be a bug in the CFQ scheduler. I found the
> following in the 2.6.28 changelog: "cfq-iosched: fix queue depth
> detection". See also
> http://www.eu.kernel.org/pub/linux/kernel/v2.6/ChangeLog-2.6.28 or
> http://lkml.org/lkml/2008/8/22/39.

Iirc it has been with RHEL5/CentOS5 2.6.18 based kernels.. 

Mike Christie has been writing about this aswell.. dunno about what kernels
he has seen it with.

Then again CFQ was designed for "single disk workstations".. 

-- Pasi

You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to open-iscsi@googlegroups.com
To unsubscribe from this group, send email to 
For more options, visit this group at http://groups.google.com/group/open-iscsi

Reply via email to