You are not logged in.

#26 2006-01-08 05:50:21

Paul
Member
Registered: 2004-04-12
Posts: 72

Re: Benchmarking Filesystems Part ||

When I compile a kernel with reiserfs, the temp of my laptop's cpu reaches mid 90's. Using JFS, it never breaches 50.

Paul

Offline

#27 2006-01-08 07:24:39

Neuro
Member
From: Poland
Registered: 2005-10-12
Posts: 352

Re: Benchmarking Filesystems Part ||

Paul wrote:

When I compile a kernel with reiserfs, the temp of my laptop's cpu reaches mid 90's. Using JFS, it never breaches 50.

Paul

Are you talking of F or C degrees?

Offline

#28 2006-01-08 15:24:34

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking Filesystems Part ||

As said before, ReiserFS is probably not a good idea for laptops, since it's very CPU-intensive.

Offline

#29 2006-01-08 19:41:16

Grady
Member
Registered: 2005-12-26
Posts: 10

Re: Benchmarking Filesystems Part ||

Would I have to reformat my partition(s) to enable DIR_INDEX? How would I do this best?

Offline

#30 2006-01-08 20:07:45

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking Filesystems Part ||

You can use tune2fs to enable dir_index, but you have to do e2fsck on those partitions afterwards. See this thread for more information.

Offline

#31 2006-01-10 16:15:54

Aletheuo
Member
Registered: 2004-03-26
Posts: 59

Re: Benchmarking Filesystems Part ||

I've been using ReiserFS exclusively but am now considering XFS, but it seems XFS in this forum is not recommended unless you have SATA or a large ATA disk, how big does the disk need to be to be happy with XFS?
Or is JFS a better alternative for 20-40gig hard disks?


Arch Linux (Duke)
JabberID:  cgill27@jabber.org
IRC: Aletheuo
Registered Linux User #354975

Offline

#32 2006-01-10 17:08:28

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking Filesystems Part ||

It's a matter of speed, not size. If you're going to use XFS for your root partition, you want to have a fast hard drive, preferably SATA or SCSI. You see, XFS performs very well with large files, but in my experience the performance with large directories containing lots of small files is absolutely abominable.

Offline

#33 2006-01-11 15:59:47

Bralkein
Member
Registered: 2004-10-26
Posts: 354

Re: Benchmarking Filesystems Part ||

I used to use ReiserFS V3, and it was fine. However, I was intrigued by the new Reiser4, so I tried it out. It seemed to be extremely fast when running pacman and also in general, but I had a pretty serious problem, whereby it wouldn't write files to HDD all the time, it was like sometimes it forgot to do it or something. Even if I tried to use sync after every write, it still didn't work, so obviously I paid the price for using a new, unstable FS :S

Resier4 was very quick at reading stuff though, so I would definitely use it again in the future when it gets merged into the mainline kernel and is a bit more stable.

After my problems with Reiser4, I decided to give Ext3 a try, because it got quite a few glowing reviews on this forum, but I am not especially happy with it. It seems about as fast as ReiserFS V3, except for when I use pacman, at which point it is unacceptably slow. Also, mount automatically forces a fsck after 33 mounts or something, which I guess I can turn off but I figure it is probably there for a reason and I don't want to corrupt my data.

If I were to reinstall Arch today, I would either go back to using ReiserFS V3, or I would give JFS a try. I don't understand all of the complaints about ReiserFS V3 mount times and data corruption, because it has always fared rather well for me. My PC isn't blazingly fast, it's a 1.3GHz Athlon, but the mount times seem negligible. Also, I have a server at my Mum's house running ReiserFS V3, and when I was changing a fuse round there the other day I ended up turning the power for the entire house off and on again about five times in half an hour, I forgot to turn the server off first, but all the data was perfectly fine.

Offline

#34 2006-01-11 21:08:39

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking Filesystems Part ||

Did you use dir_index when creating your ext3 filesystem? If you did not, then you have the answer to your pacman problems.

Offline

#35 2006-01-12 17:49:44

Bralkein
Member
Registered: 2004-10-26
Posts: 354

Re: Benchmarking Filesystems Part ||

Nah, that must be it... TBH, it's not really that much of a big deal, I can't really be arsed with changing it to dir_index... I think that once Reiser4 stabilises out a bit I will go back to using that though. I basically have a soft spot for insane, forward-thinking visionary projects, even if they don't turn out to be quite as amazing as the dream suggested... they're just so awesome!

Offline

#36 2006-01-12 20:09:58

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking Filesystems Part ||

Then maybe you should try out ZFS when it gets ported. :twisted:

Anyway, FWIW, you DON'T have to reformat in order to enable dir_index -  check out this Arch Wiki page for more information. You can use tune2fs to enable dir_index on an UNMOUNTED filesystem like so:

# tune2fs -O dir_index /dev/hdXY

Then you have to do a fsck in order for it to take effect:

# e2fsck -D -f /dev/hdXY

(The -D option is the relevent one there, as it tells the filesystem to reindex all of the directories.)

Note that this procedure could take a good bit longer than your average fsck.

Bralkein wrote:

Also, mount automatically forces a fsck after 33 mounts or something, which I guess I can turn off but I figure it is probably there for a reason and I don't want to corrupt my data.

You can indeed turn it off without causing any problems, as per the Wiki page. Ext3 is a journalled filesystem, and therefore does not need to be fscked everytime your machine is shut down improperly; as long as the "pass" entry for your ext3 filesystems in /etc/fstab is greater than 0, you don't really have to worry.

(Ext2 is another matter entirely; it's nonjournalled, so it's best to leave it as it is...)

Offline

#37 2006-01-13 18:30:21

Bralkein
Member
Registered: 2004-10-26
Posts: 354

Re: Benchmarking Filesystems Part ||

Cheers! Yeah I know you don't need to remount it, I am just very lazy... I will probably sort it out at some point  :oops:

That was the thing that confused me about the fsck every 33 mounts... I know that ext3 is basically just ext2 with a journal glued on the top, so it seemed very confusing that the default settings should be so fsck-happy. Why on Earth did whoever it was decide to make it do that?

ZFS looks kind of interesting. I have not looked into it particularly deeply, but it seems to be more of a server/business-oriented FS to me. I'm sure it would come in very useful if I had to run a big fileserver, but for Reiser4, the features look much more impressive for desktop use.

Apparently, certain features of Reiser4 (I think it is support for file metadata or something) mean that sometime in the future, it will be possible to basically have desktop search capabilities that hook directly into the filesystem, which is supposed to be a much more optimal solution than current methods employed by Apple's Spotlight, Google Desktop Search, etc. In fact, you may know that this is what Microsoft were originally planning to do with WinFS, but they never delivered on that promise. WinFS should be shipping with Windows Vista, but they have broken all their promises and I think it's just supposed to be an equivalent of Spotlight now... no longer a full FS, just a desktop search framework, with NTFS still managing all of the files in the background.

Offline

Board footer

Powered by FluxBB