On Sat, Oct 05, 2002 at 04:36:29PM -0700, Joseph Carter wrote:
>Journals actually decrease performance a marginal amount.  It's not very
>much though.

That's not entirely true.  In ext2, for file-system meta data updates
there's this delicate little dance where you have to update different
parts of the file-system then ensure that they're committed to disc
before going on and updating other parts, to ensure that the filesystem
is not left in a state where a crash would totally munge things.

With the journal, you can stream the journal to disc, ensure that it's
committed, then go on and do all the other updates in whateer order
makes sense, then do another commit at the end of it.  That can be a big
performance gain.

It's also not uncommon to put the journal on NVRAM devices so that the
journal is super fast.  That's another way they increase performance of
journaled filesystems.

If you are running SMP machines with quotas enabled you probably don't
want to run ext3 right now.  I prefer ext3 on my laptop though, because
in the past we had problems with our laptops and reiser -- every one of
our laptops that was running reiserfs had at least a minor data
corruption issue with reiserfs, and one of them had a major (wipe and
start over) problem during a year of use.

Sean
-- 
 Reading computer manuals without the hardware is as frustrating as reading
 sex manuals without the software.  -- Arthur C. Clarke
Sean Reifschneider, Inimitably Superfluous <[EMAIL PROTECTED]>
tummy.com, ltd. - Linux Consulting since 1995.  Qmail, Python, SysAdmin
_______________________________________________
Eug-LUG mailing list
[EMAIL PROTECTED]
http://mailman.efn.org/cgi-bin/listinfo/eug-lug

Reply via email to