O&O Defrag for Linux...

shiny_spoon · 11-14-2003, 03:21 AM

Hey everyone!

While speaking with one of my friends recently, he mentioned he had stumbled upon a product called O&O Defrag Linux. You guessed it, it's a third-party disk defragmenter. But for Linux??

I was always under the impression that EXT2/EXT3 file system's didn't suffer from fragmentation of files unless there was a severe lack of space on the drive.

With that having been said, I was wondering if any of you had the chance to try this thing. And if so, were the results good/bad, any complications...? Or maybe comments you might have about it?

I'm interested in trying it out myself, but having recently almost lost 120GB of data resizing partitions, I'm not too keen on the possibility of data loss...

Thanks!

DirtDart · 11-14-2003, 05:29 AM

Not quite sure what they folks over at O&O are smoking, but they're trying to sell a product that isn't needed.

O&O Defrag is a great product. For Windows. The Windows (any version) defrag utility blows goats, so a 3rd party product (O&O in this case) is usually needed.

In linux, there is no need for it. Even if you're system is shut down improperly (power loss, you accidently shut it off without shuting linux down, ect), the niffy, and included, utility e2fsck. e2fsck works just fine; no need in using another utility.

shiny_spoon · 11-14-2003, 12:53 PM

Quote:

Originally posted by DirtDart
Not quite sure what they folks over at O&O are smoking, but they're trying to sell a product that isn't needed.

There you go, that's exactly what I was thinking. But I read an article where the people at O&O stated a performance increase of up to 10%... How's that even possible?

I'm still pretty skeptical about this thing.

stickman · 11-14-2003, 03:08 PM

I wonder which type of fragmentation it fixes.

Tinkster · 11-14-2003, 04:27 PM

Well, on boot ext2 (ext3 too?) will tell you
about the percentage of noncontiguous
blocks ... I've never seen more than 2%
fragmentation in any linux installation :)

To me, that's good enough, and I don't
think that their test-scenario is applicable
to the real-world ;) ....

I mean, to fragment a linux filesystem
you'd have to first fill it to the brim with
small files, delete hepas of non-adjacent
ones, and then write files that don't fit in
any gaps.

Not a very "normal" behaviour for a linux
system.

Cheers,
Tink

DirtDart · 11-15-2003, 03:21 AM

Quote:

I've never seen more than 2%
fragmentation in any linux installation

I've never seen it above 2% either.

narendra_i · 11-15-2003, 08:50 AM

i have never bothered about it

voodooutt · 11-20-2003, 06:55 PM

so, what people are saying here.... there really IS NO NEED to defragment a linux partition?

even after a looooong time?

im n00b... so bear with me

dalek · 11-20-2003, 07:16 PM

Nope, no need to worry about defrag. It takes care of itself. I actually saw mine get to about 4% a while back and when I rebooted, it did the file system check and fixed the errors. I then rebooted with the fsck option and it was at about .1%.

I wouldn't worry about it. Linux is so efficient, you won't notice it anyway.

Later

mganss · 11-21-2003, 06:10 AM

Hi,

I'm with O&O Software and would like to comment on the issue of fragmentation on Linux filesystems, specifically ext2 and ext3:

It is a myth that ext2/ext3 does not fragment. Every file system I know of fragments. A few years ago it was also claimed that NTFS did not fragment, which of course it does (see http://www.oo-software.com/en/produc...aq_pro.html#a3)

If you think about it, how can a filesystem that stores its data in blocks on a hard disk not fragment? If it had some sort of logic that dynamically reordered blocks on the disk in order to prevent fragmentation (somewhat like a built-in defragmenter), that would reduce performance of the file system.

Tim Robbins of Gentoo says this in one of his file system articles on IBM developerworks:
"...any often-modified ext2 filesystem will gradually get more and more fragmented, and thus slower. Even worse, there are no production-quality ext2 filesystem defragmenting programs currently available." (see http://www-106.ibm.com/developerwork...titiontip.html)

As for our performance stats: They were carried out using Intel iometer (now an Open Source project at http://www.iometer.org/) and showed a consistent performance increase of 10-15%.
Keep in mind that fsck only shows file fragmentation, i.e. the ratio of fragmented files to non-fragmented files. Block fragmentation, i.e. the ratio of blocks in fragments to total number of blocks, is more relevant to the performance of the file system. You can find out block fragmentation with the "-a" switch of O&O Defrag Linux.

DirtDart: fsck does nothing against fragmentation, it checks a file system against errors and tries to fix them, much like chkdsk on Windows.

Bear in mind that O&O Defrag Linux is still beta software and should not be used on production systems.

Regards

DirtDart · 11-21-2003, 09:07 AM

Quote:

DirtDart: fsck does nothing against fragmentation, it checks a file system against errors and tries to fix them, much like chkdsk on Windows.

Opps...my bad.

Not knowing a whole bunch about disk fragmentation in linux, I won't put much info out here, but I do know that I've never seen anything out of the single digit range. Of course, I'm not dealing with production servers, just home PCs, so the amount of fragmentation might be a little different.

If the software increases performance as much as you say it does, it should do very well. I know your Windows version kicks some serious ass, so you have your proven track record going for you.

shiny_spoon · 11-21-2003, 01:53 PM

Thanks for clearing things up a bit mganss. I found your post quite useful.

sueisfine · 02-24-2004, 05:24 AM

Quote:

Originally posted by DirtDart
I've never seen it above 2% either.

what does this non-contigous mean? like un-defragmented? i have :

Code:

/dev/hdb7: 2543/8552448 files (1.1% non-contiguous), 3686661/17089135 blocks

and i just copyed those files there...

im having serious problems with my /dev/hdb drive and reiserfs. so i will go to ext3...

jago25_98 · 02-24-2004, 03:56 PM

Thanks to the O&O guy, this has cleared up a long held confusion for me

sueisfine · 03-28-2004, 04:41 PM

Code:

bash-2.05b# fsck.ext3 -fv /dev/hdb6
e2fsck 1.34 (25-Jul-2003)
Pass 1: Checking inodes, blocks, and sizes
Pass 2: Checking directory structure                                           
Pass 3: Checking directory connectivity                                        
Pass 4: Checking reference counts
Pass 5: Checking group summary information                                     
                                                                               
    3691 inodes used (0%)
     343 non-contiguous inodes (9.3%)
         # of inodes with ind/dind/tind blocks: 3115/1982/0
 4919423 blocks used (31%)
       0 bad blocks
       0 large files

    3324 regular files
     358 directories
       0 character device files
       0 block device files
       0 fifos
       0 links
       0 symbolic links (0 fast symbolic links)
       0 sockets
--------
    3682 files

partition is for my mp3's.

there are non-contiguous inodes (9.3%)? is that normal?