2.6.24.7-rt8

2.6.24.7-rt8

[Steven Rostedt]

We are pleased to announce the 2.6.24.7-rt8 tree, which can be
downloaded from the location:

  http://rt.et.redhat.com/download/

Information on the RT patch can be found at:

  http://rt.wiki.kernel.org/index.php/Main_Page

Changes since 2.6.24.7-rt7

 - Read Write locks with multiple readers (Steven Rostedt)

 The previous versions of the RT tree would serialize the RW locks
 for all readers to allow for priority inheritance to take place.
 This change allows for multiple readers but still maintain priority
 inheritance. The number of readers that can hold a lock at any time
 may be limited by

   sysctl kernel.rwlocks_reader_limit

 Setting this to <= 0 will give allow for any number of readers. The
 default is set to NR_CPUS.


 ** Note: I made this change separate to allow for easy testing in case 
          bugs have been introduced by this change.

to build a 2.6.24.7-rt8 tree, the following patches should be applied:

  http://www.kernel.org/pub/linux/kernel/v2.6/linux-2.6.24.tar.bz2
  http://kernel.org/pub/linux/kernel/v2.6/patch-2.6.24.7.bz2
  http://rt.et.redhat.com/download/patch-2.6.24.7-rt8.bz2



And like always, my RT version of Matt Mackall's ketchup will get this
for you nicely:

  http://people.redhat.com/srostedt/rt/tools/ketchup-0.9.8-rt3


The broken out patches are also available.



-- Steve

Re: 2.6.24.7-rt8

[Mike Galbraith]

Hi Steven,

On Tue, 2008-05-20 at 00:46 -0400, Steven Rostedt wrote:
> We are pleased to announce the 2.6.24.7-rt8 tree, which can be
> downloaded from the location:
> 
>   http://rt.et.redhat.com/download/

I ran mysql+oltp (ro) on 2.6.24.7-rt7 and 2.6.25-rt2 to compare
throughput to their parent trees, and ended up with strange (and
verified) looking performance curves for both trees.  The changes in
2.6.24.7-rt8 and 2.6.25.4-rt3 cured the primary oddities.

(it looks to me like there may be a little load-balancing issue, peak
throughput isn't where it should be for my Q6600 box)

	-Mike

Re: 2.6.24.7-rt8

[Mike Galbraith]

[-- Attachment #1: Type: text/plain, Size: 734 bytes --]

Ahem.  Attach the data _before_ poking xmit.

On Wed, 2008-05-21 at 08:25 +0200, Mike Galbraith wrote:
> Hi Steven,
> 
> On Tue, 2008-05-20 at 00:46 -0400, Steven Rostedt wrote:
> > We are pleased to announce the 2.6.24.7-rt8 tree, which can be
> > downloaded from the location:
> > 
> >   http://rt.et.redhat.com/download/
> 
> I ran mysql+oltp (ro) on 2.6.24.7-rt7 and 2.6.25-rt2 to compare
> throughput to their parent trees, and ended up with strange (and
> verified) looking performance curves for both trees.  The changes in
> 2.6.24.7-rt8 and 2.6.25.4-rt3 cured the primary oddities.
> 
> (it looks to me like there may be a little load-balancing issue, peak
> throughput isn't where it should be for my Q6600 box)
> 
> 	-Mike

[-- Attachment #2: zz.pdf --]
[-- Type: application/pdf, Size: 16993 bytes --]

Re: 2.6.24.7-rt8

[Steven Rostedt]

On Wed, 21 May 2008, Mike Galbraith wrote:
> On Tue, 2008-05-20 at 00:46 -0400, Steven Rostedt wrote:
> > We are pleased to announce the 2.6.24.7-rt8 tree, which can be
> > downloaded from the location:
> >
> >   http://rt.et.redhat.com/download/
>
> I ran mysql+oltp (ro) on 2.6.24.7-rt7 and 2.6.25-rt2 to compare
> throughput to their parent trees, and ended up with strange (and
> verified) looking performance curves for both trees.  The changes in
> 2.6.24.7-rt8 and 2.6.25.4-rt3 cured the primary oddities.

The difference between 24-rt7 and 24-rt8 as well as with 25-rt2 and 25-rt3
is the deserialization of rwlocks. The rt patch as always serialized
readers to allow for priority inheritance with rwlocks. I wrote a patch to
break that and allow for multiple readers and still keep the priority
inheritance.

Your test probably had a bottle neck on a rwlock (same for rwsems)
somewhere that the new code solved.

-- Steve