Feature Request

[Samuel Flory]

Dax Kelson wrote:

>On Thu, 2003-07-24 at 16:10, Samuel Flory wrote:
>  
>
>>  I've never seen a benchmark were ext3 won.  I'd love someone to show 
>>me one that's not just a single narrow test.
>>    
>>
>
>In general, yes, but you must not have looked too hard.
>

  No you didn't look hard enough.

>
>http://www.gurulabs.com/ext3-reiserfs.html
>  
>

  Ext3 writeback beats reiserfs in the benchmarks but ordered the mode 
that Red Hat use by default doesn't get high marks.  Also if you look at 
the test the minute ext3 ordered actually starts writting to disk it 
gets it's ass kicked.  Under heavy load reiserfs kicks ass all over the 
place


http://www.gurulabs.com/ext3-reiserfs-3.html
"This first test uses relatively small files in the range of 1000 bytes 
to 9000 bytes, and has a net result in causing about 150MB of data to be 
read and written to. This should fit well within the RAM on the system. 
This should be interesting to see how well the journaling filesystems do 
when there are lots of operations mostly in memory."

  Cool ext3 is fast if I'm not actually writing to disk!!


http://www.gurulabs.com/ext3-reiserfs-5.html
"This last test doubles the amount of files to 4000, and drastically
bumps the maximum file size to 300,000 bytes, and has a net result in
causing about 19GB of data to be read and written to. This by far
exceeds the RAM on the system and should generate a massive amount of
IO on the drives. This should be interesting to see how the filesystems
compare when the system is being crushed."

This what you care about in the real world!!

-- 
Once you have their hardware. Never give it back.
(The First Rule of Hardware Acquisition)
Sam Flory  <sflory at rackable.com>