Re: Mount options for Optimizing ext2/ext3 performance with Maildir's
On Monday 18 October 2004 18:15, Ian Forbes wrote:
> I have a mailserver with load average sitting somewhere between 1 and 2. It
> is running exim serving a couple of thousand Maildir mailboxes and also has
> a bunch of antivirus / antispam programs running on it. It has a pair of
> ide hard drives running mirrored, raid1. I really do not want to start on
> reiser or XFS. Reliability is my major concern and I do not think they are
> warrented. But I would like to maximise my performance with the existing
> ext2/ext3 partions.
Well, reliability is relative. You can have a near-perfect reliable ext2 FS,
but what is that reliability worth if it means that at every crash the box
will be down for 5 hours doing its e2fsck on the 200 GB mail volume...?
> Is ext3 faster or slower than ext2?
I'm no expert myself but I've read that ext2 is faster than ext3 because it
has no additional overhead caused by the journaling functions. But see above
what that tradeoff can result in. Journaled filesystems are here for a good
reason (not to mention filesystem consistency).
Everyone has to live with his/her own choices and the consequences thereof, so
I will not try to convince you to run XFS or reiser, but in my own experience
reiserfs has proven more reliable than ext3.
Other than that, I've heard that reiserfs is especially efficient with small
files, and XFS being especially efficient for big files. I have no data on
ext3, but the web is littered with benchmarks (all with different results of
course, since benchmarks are very rarely neutral ;-)
> Currently I have everything in one big root partition. If I mount it with
> "noatime" will a hole bunch of things stop working, like the automatic
> reloading of files in /etc/cron.d/ ?
Don't know, but you seriously have to consider making different partitions for
the various mountpoints, not only for the atime issue but for added security.
I'm not talking one / and one mailspool, but giving at least /tmp, /usr, /var
and /home (when applicable) their own partitions too. This is good for a lot
of reasons; with FS crashes you do not stand to lose everything (as you do
with 1 single partition), you make the likelyhood of a DoS much smaller, and
you can set specific mount options per drive, like noexec on /home, or nosuid
on /tmp, or whatever you desire, which wholly depends on the role of the
machine and its users' granted rights.
> The system is running with a 2.4.18. Is there anything to be gained from
> upgrading to a later 2.4 or a 2.6 kernel.
Isn't 2.4.18 a tad old by now ? Are you security-conscious ?