Spreading the word

[sag47]

I've been following and using Jeroen's defrag software since it was early open source (JkDefrag).  And still use and enjoy MyDefrag today.  I still wish it was open source but I understand reasoning behind closing source code.

Basically I am new to the forum but not new to the software and have written an article on another forum's tech section that I frequent promoting MyDefrag.

Please review it and alert me of any errors or inconsistencies in my article so that I may correct them.  I want it to be as correct as I can be.  Also please understand that I wrote this article geared towards non-technical people so that they may benefit from it as well as advanced users.  So I will not modify terminology if I feel that it will complicate contemplation for the basic user.  But please still alert me of them if you feel it should be changed!

Here's the article and you can refer to it as many times as you like.

http://www.clubbleach.org/forums/showthread.php?p=4156699

SAM

[jeroen]

I have read your article. Here are my remarks, you asked for it!

- Fragmentation on Linux is not less likely than on Windows. This is an often quoted myth.
- In the beginning of your article you only mention defragmentation. This however is only a small part of MyDefrag, it's the optimization (organizing files on the disk) that is actually the most important.
- The innermost ring on the disk is the slowest part of the harddisk, the outermost is the fastest. Not as your graph shows.
- I recommend to run the daily script every day. If there is little work then MyDefrag will finish quickly.

[sag47]

- Fragmentation on Linux is not less likely than on Windows. This is an often quoted myth.

How does Linux handle fragmentation then?  It must do something to stay reliable and fast.  I hear about Linux filesystems (ext3 specifically) handling fragmentation on the fly though I just see people quoting it and not official documentation references on the matter.  Is this true?

- In the beginning of your article you only mention defragmentation. This however is only a small part of MyDefrag, it's the optimization (organizing files on the disk) that is actually the most important.

I'll go back and edit the article highlighting optimization with more emphasis.

- The innermost ring on the disk is the slowest part of the harddisk, the outermost is the fastest. Not as your graph shows.

Makes sense now that I think about it.  Since the inner ring and outer ring don't separate like cyclists on a track.  Thanks for pointing that out.  So then would it be correct to say that MyDefrag moves data to the outer rings of the disks then or does it do something different?

- I recommend to run the daily script every day. If there is little work then MyDefrag will finish quickly.

What is the point of running the monthly script then?  Should I run only the daily optimization?

Thanks for getting back to my post.

[Darlis]

- Fragmentation on Linux is not less likely than on Windows. This is an often quoted myth.

How does Linux handle fragmentation then?  It must do something to stay reliable and fast.  I hear about Linux filesystems (ext3 specifically) handling fragmentation on the fly though I just see people quoting it and not official documentation references on the matter.  Is this true?

ext4 has a build in defragmentation feature.  http://en.wikipedia.org/wiki/Ext3   http://en.wikipedia.org/wiki/Ext4

- I recommend to run the daily script every day. If there is little work then MyDefrag will finish quickly.

What is the point of running the monthly script then?  Should I run only the daily optimization?

http://www.mydefrag.com/FAQUsing-WhatIsTheBestDefragmentationoptimizationStrategyForMe.html

[quanthero]

Most Linux filesystems avoid file fragmentation quite well as long as a partition stays relatively empty. This is achieved by reserving space for files, using delayed allocation, etc. However, these techniques fragment free space very quickly. Fragmented free space means that filesystem runs out of contiguous space to write new files, so newly written files become fragmented. So, performance stays high as long as partition is relatively empty.
For these reasons, I think we really need a good optimization program for Linux, which will not only defragment files but will consolidate free space into as few chunks as possible.

*Note: I use this tool to determine file fragmentation on Linux FSs. I also found that Paragon Total Defrag is the only tool that shows bitmap of ext* FSs which is helpful for determining the degree of free space fragmentation.

[TheAnonymous]

My experinces are written below:

The innermost ring should be the fastest for seeking, the outermost ring has a greater velocitry and produces better read/write speeds. The outermost ring might be faster overall for drives that are 5,400RPM or lower -- but for a drive that is more than, say, 30mb/s from the innermost ring, won't benefit from the increased read speed from the outermost ring, only the write speed. The average non-server use will never have their disk actually doing more than around 30mb/s of disk reads at a single time, especially if they have a lot of memory and hence data can from memory more often as it would be placed in the page cache.

Most Linux file systems don't get file fragmentation untill it is 90% full, but they still have free space fragmentation.

[poutnik]

My experinces are written below:
The innermost ring should be the fastest for seeking....

why should innermost ring be faster for seeking ?
Is disk head track stepping / settling faster for inner tracks ?
Even if it is, they are far from center of disk activity and disk head trevelling to inner tracks takes longer.

[cheese]

there is lot's of false assumptions out and frequently things are more complicated than what they appear to be ...

Dominic Humphries wrote an intro some time ago:
http://geekblog.oneandoneis2.org/index.php/2006/08/17/why_doesn_t_linux_need_defragmenting
http://geekblog.oneandoneis2.org/index.php/2008/11/13/fighting-fragmentation-on-linux
there's plenty of useful information in the comments if you can spare the time...

[jeroen]

Dominic Humphries wrote an intro some time ago:

I wish that article was removed from the internet. It has been quoted many times here on the MyDefrag forum, as if it was gospel. It's a clever explanation of a difference that does not exist, assumes things that do not happen, and is totally theoretical and contains zero hard information. All the article does is keeping the myth alive about Linux not needing defragmentation. And the second article is downright wrong, as anybody playing a fragmented movie on a busy machine will know.

[TheAnonymous]

Disregard my last statement. I did a test with HDTune 4.0 and the outer-ring is indeed the fastest overall, for seek, read and write times.

The interesting thing is that the first 40GB of my hard drive (which is what I have it short stroked to) has an access time of 4.5MS, where across the whole drive is 7.1MS. That's a pretty signifcant difference.

Fragmentation is just a file split across the disk, if the file is defragmented but at the end of the disk, it will still be accessed much slower.

Filesystems like EXT4 might allocate files so that the files do not get fragmented until the drive is about 90% full, but they still have free space fragmentation and the order of files is still suboptimal.

Jeron has pointed out many times that MyDefrag is about more than just defragmentation. And that the only impact to disk performance is not just due to file fragmentation.

No offense, cheese, but that second article you posted is just terrible! Especially since it only talks about large files and not, let's say, <50MB files that are usually accessed more often. Although someone with 2GB or more of memory will have smaller files be placed into the page cache anyway, they will still have to be the read from the disk the first time they are accessed each session (each usage without an OS reboot).

Has anyone read the comments on the first article posted? There are claims that fragmentation can help performance or defragmenting files on Linux filesystems can degrade performance. Unless the defragmentation process moves them farther from the beginning of the disk, where is the sense in such claims?

I agree with the original post, the word of MyDefrag really does need to be spread. I recently tried out the "SMARTPlacement" feature in a commercial defragmenter named PerfectDisk. I analyzed what it did to my disk and it did this:

Moved all files in WinSXS to the BEGINNING of my disk, along with a bunch of other spacehogs.
Small files were farthest from the beginning of the disk, and had tons of free space gaps inbetween them.

I think everyone can realize what is wrong with the above.

MyDefrag is the only defragmentation program I'll ever use