Re: 2.5.70-mm1

[Andrew Morton]

"Martin J. Bligh" <[e-mail]> wrote:
>
> > SDET 128  (see disclaimer)
> >                            Throughput    Std. Dev
> >                2.5.66-mm2       100.0%         0.6%
> >           2.5.66-mm2-ext3         3.9%         0.4%
> > 
> > SDET 128  (see disclaimer)
> >                            Throughput    Std. Dev
> >           2.5.70-mm1-ext2       100.0%         0.1%
> >           2.5.70-mm1-ext3        22.7%         2.0%
> 
> Andrew pointed out I should turn off extended attributes, I reran
> like this ... not much change (with error margins)
> 
>  SDET 32  (see disclaimer)
>                            Throughput    Std. Dev
>           2.5.70-mm1-ext2       100.0%         0.2%
>           2.5.70-mm1-ext3        30.9%         7.8%
>           2.5.70-mm1-noxa        34.6%         6.5%

OK, a 10x improvement isn't too bad.  I'm hoping the gap between ext2 and
ext3 is mainly idle time and not spinning-on-locks time.

> 
>    2024927   267.3% total
>    1677960   472.8% default_idle
>     116350     0.0% .text.lock.transaction
>      42783     0.0% do_get_write_access
>      40293     0.0% journal_dirty_metadata
>      34251  6414.0% __down
>      27867  9166.8% .text.lock.attr

Bah.  In inode_setattr(), move the mark_inode_dirty() outside
lock_kernel().

>      20016  2619.9% __wake_up
>      19632   927.4% schedule
>      12204     0.0% .text.lock.sched
>      12128     0.0% start_this_handle
>      10011     0.0% journal_add_journal_head

hm, lots of context switches still.