Re: jbd/jbd2 performance improvements

[Eric Sandeen <sandeen@redhat.com>]

Ric Wheeler wrote:
> Solofo.Ramangalahy@bull.net wrote:
>> Hi Ric,
>>
>>   
>>>>>>> On Wed, 15 Oct 2008 13:29:55 -0400, Ric Wheeler <rwheeler@redhat.com> said:
>>>>>>>             
>>     Ric> We are going to try and poke at this - do you suspect a single or 
>>     Ric> multi-threaded test would work best?
>>
>> I've performed some tests:
>> http://www.bullopensource.org/ext4/20081013-2.6.27-rc9-ext4-1-akpm-fix-run6/
>> http://www.bullopensource.org/ext4/20081013-2.6.27-rc9-ext4-1-akpm-fix-run6/results_sorted.txt.html
>>
>> I now realize that the results may not be valid since I used kvm, but
>> they do show variation wrt. the number of threads.
>>
>> So you may want to test both single and multi-threaded.
>>
>>   
> A very thorough test, but the results don't seem to point to a 
> consistent winner.
> 
> I agree that running without KVM in the picture might be very 
> interesting. Eric has some similar tests underway, I think that his 
> results were also inconclusive so far...

Yep, I've yet to find an fs_mark invocation, at least, which shows a
clear winner.  I also ran w/ akpm's suggested io_schedule watcher patch
and never see us waiting on this lock (I did set it to 1s though, which
is probably too long for my storage).

-Eric