ext2fs slow?

Barry barry at hartford.uconn.edu
Thu Dec 28 09:09:54 PST 2000

Matthias Benkmann wrote:
> I got a new hard disk for Christmas and moved my LFS system to a 13GB
> partition on the new hard disk. My problem is now that while mounting this
> partition in read-only mode is fast, mounting it read-write takes many
> seconds, at least when done the first time. This slows down booting a lot.
> Is this an ext2fs problem? Mounting the vfat partition of the same size on
> the same disk is fast.
> I've seen ReiserFS mentioned around here several times. Is that one
> mounted faster (I'm not concerned about faster fschks. I know that
> ReiserFS is faster in that respect) ?


OK, this is a two-fold answer...

There are two problems here.  Yes, ext2fs is one of them, and yes -
reiserfs scales better to larger file systems.  I've tested it. 
However, the kernel also has something to do with it.  Older versions of
the Linux kernel start to scale down the loading of the ext2fs when you
get beyond something like 3 or 4 gigs.  Maybe even less on some
systems.  Superior support for large file systems will be in the 2.4

Now, as for reiserfs.  My tests have been very promising.  Not only will
you see an increase in speed, your file system will also survive
corruption in a superior fashion.  I've tested this and found that
reiserfs seems to not corrupt anything that isn't being directly written
to.  And the corruption is isolated to that file.  Viva journaling. 
Also, for the same reasons, you'll notice a considerable space saving. 
I placed a full  Red Hat 6.2 install 8 monthes ago onto the reiserfs
type of that time.  The initial usage of the partition was about 1.5
gigs in ext2.  Upon copying (with cp -av *) the files over to the
reiserfs partition and booting, I saw an instant space savings of over
300 megs... that's something like 20% of the drive was saved.  My tests
were with the older tea hash function.  Now, reiserfs gives a choice of
the r5 and tea hash functions.  To my knowledge, r5 is supposed to scale
better, but I haven't tested it extensively.

hope that helped :)

Unsubscribe: send email to lfs-discuss-request at linuxfromscratch.org
and put unsubscribe in the subject header of the message

More information about the lfs-dev mailing list