Feature Request
Benjamin Vander Jagt
benjaminvanderjagt at adelphia.net
Thu Jul 24 22:46:14 UTC 2003
Aye!
I switched to ReiserFS, because my brother and I simply couldn't play
Quake3 or Unreal Tournament 2003 when every ten minutes, we'd get slowed
to nearly a halt while the hard drive goes nuts.
Plus, I used to get a lot of power outages. Ext3 has, a few times, been
totally destroyed (even fsck doesn't find data) where I have *never*
lost *any* data to ReiserFS, even on two test hard drives that were
loaded with bad sectors. When I switched to ReiserFS, I did so for the
performance, believing all the rhetoric that it wasn't as reliable as
Ext3.
It seems to me that more of the filesystem problems I've seen forum
posts about on the internet have to do with Ext3 than ReiserFS.
Of the top 10 distros in Distrowatch, only 4 do not support ReiserFS
(and several prefer ReiserFS). Those 4 are Red Hat, Knoppix, Lycoris,
and Morphix. Knoppix and Morphix obviously don't have a wide range of
supported FS's, because they are live CDs, and Lycoris is silly in every
other way anyway.
Respectfully speaking, Red Hat's ReiserFS information is very old. I
believe the Kernel changelog (I think for 2.4.18, but I'm not sure)
talks about a coding error that made ReiserFS unreliable.
The problem I have is that I have extensively used ReiserFS, and unless
Anaconda supports it (such as installing over a previous installation),
then I just won't be able to use the new Red Hat without a big backup
procedure.
Anyway, I'm tipping out the door. I'll just stop the mailing to my box,
so I can still sorta respond to the archives.
On Thu, 2003-07-24 at 17:15, Samuel Flory wrote:
> Ethan Bonick wrote:
> I for one am not okay with ext3 as the default, and only journalling
> files system. Let's face it the default ext3 journaling mode is
> extremely slow on writes. As you can see below reiserfs is x2 as fast
> on writes, and x1.5 as fast on reads. Writeback mode isn't much
> better. And ext3 is even worse on random file system access.
>
> Reiserfs:
> + dd if=/dev/zero of=bigfile bs=1024000 count=2048
> 2048+0 records in
> 2048+0 records out
>
> real 1m16.632s
> user 0m0.000s
> sys 0m12.580s
> + sync
>
> real 0m24.649s
> user 0m0.000s
> sys 0m0.320s
> + dd of=/dev/null if=bigfile bs=1024000 count=2048
> 2048+0 records in
> 2048+0 records out
>
> real 0m36.065s
> user 0m0.010s
> sys 0m10.690s
> + sync
> + du -sh bigfile
> 2.0G bigfile
>
>
> ext3:
> + dd if=/dev/zero of=bigfile bs=1024000 count=3072
> 3072+0 records in
> 3072+0 records out
>
> real 2m15.430s
> user 0m0.010s
> sys 0m20.900s
> + sync
>
> real 0m32.166s
> user 0m0.000s
> sys 0m0.190s
> + dd of=/dev/null if=bigfile bs=1024000
> 3072+0 records in
> 3072+0 records out
>
> real 0m54.577s
> user 0m0.010s
> sys 0m16.070s
>
>
> ext2 in writeback mode
> + dd if=/dev/zero of=bigfile bs=1024000 count=3072
> 3072+0 records in
> 3072+0 records out
>
> real 2m12.781s
> user 0m0.020s
> sys 0m20.080s
> + sync
>
> real 0m25.971s
> user 0m0.000s
> sys 0m0.300s
> + dd of=/dev/null if=bigfile bs=1024000
> 3072+0 records in
> 3072+0 records out
>
> real 0m55.136s
> user 0m0.020s
> sys 0m15.430s
> + sync
>
> real 0m0.011s
> user 0m0.000s
> sys 0m0.000s
> + du -sh bigfile
> 3.0G bigfile
--
Benjamin Vander Jagt <benjaminvanderjagt at adelphia.net>
More information about the fedora-test-list
mailing list