Re: ext3 performance issue with a Berkeley db application

[Greg Louis <glouis dynamicro on ca>]

On 20030204 (Tue) at 0212:14 +0100, Matthias Andree wrote:
> 
> > My recommendation is to work out why the application is performing so much
> > overwriting, and make it stop.  
> 
> Yup, make ext3 on IDE stop this high write rate. =:->
> 
> Seriously, as long as ext3 + IDE is a problem and ext2 + IDE isn't (with
> 2.4 at least), reiserfs + IDE isn't, ext3 + SCSI isn't, there's no
> compelling reason to change the application code.

I get a different result: the comparison I sent initially among
ordered/journal/writeback was done on a _SCSI_ drive.  Building with
bogofilter from randlist.txt with data=journal takes my laptop/IDE 27
minutes and my tower/SCSI 26 and a half.  So I would be inclined to
doubt that the solution may reside in anything IDE-related.

On my laptop, simbf takes 8m 14s (ext3, data=ordered).  On the SCSI
drive mentioned above, with ext3, data=writeback, 18 seconds flat.
With ext3, data=journal, it was 3/4 done when I executed
while true; do kill -9 (simbf process); done -- the only way -- at ten
minutes; it had completely locked up access to the drive and wasn't
interested in ctrl-C.  With ext3, data=ordered, simbf completed in 7
minutes and 8 seconds.  With ext2, 13.7 seconds.  Again, this was a
SCSI drive.

Sorry I didn't get to this last night.

-- 
| G r e g  L o u i s          | gpg public key:      |
|   http://www.bgl.nu/~glouis |   finger greg bgl nu |
| Help free our mailboxes. Include                   |
|        http://wecanstopspam.org in your signature. |