一个Linux下的文件系统碎片整理工具

芬达7402

原帖由 nntp 于 2008-9-3 12:18 发表
不需要. 我很负责任的告诉你.

版主能不能说说为什么呢？

aleng

任何分区格式都会产生碎片的，这是公理。
任何分区格式都要把磁盘分块，块太大，碎片少，但浪费了空间，块太小，不整理浪费时间，整理也浪费时间，总之块小了省空间浪费时间，这是公理！

linux unix 是从来没有整理工具的，也不想有，这是定理。

为了减少碎片，reserfs用了二叉树，这个东东可以极大地减少碎片，磁盘越大，小文件越多，越好些

为了减少碎片，ext3 使用了大文件存储块，极大地浪费了空间，但90%不需要整理碎片！

ext4 将开始着手解决这些问题。

可以说 ext3 绝不是 不会产生碎片的神话，而是非常简陋！！！

ntfs用了极小的存储块 默认4k 极大地节省了空间，但有碎片，极大地浪费了时间。

[ 本帖最后由 aleng 于 2008-9-9 15:22 编辑 ]

nntp

原帖由 芬达7402 于 2008-9-7 09:30 发表

版主能不能说说为什么呢？

. . . That is a question that crops up with regularity on Linux forums when new users are unable to find the defrag tool on their shiny new desktop. Here's my attempt at giving a simple, non-technical answer as to why some filesystems suffer more from fragmenting than others.

Rather than simply stumble through lots of dry technical explanations, I'm opting to consider that an ASCII picture is worth a thousand words. Here, therefore, is the picture I shall be using to explain the whole thing:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
b  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
n  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

This is a representation of a (very small) hard drive, as yet completely empty - Hence all the zeros. The a-z's at the top and the left side of the grid are used to locate each individual byte of data: The top left is aa, top right is za, and bottom left is az. You get the idea, I'm sure. . .

We shall begin with a simple filesystem of a sort that most users are familiar with: One that will need defragmenting occasionally. Such filesystems, which include FAT, remain important to both Windows and Linux users: if only for USB flash drives, FAT is still widely used - unfortunately, it suffers badly from fragmentation.

We add a file to our filesystem, and our hard drive now looks like this:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t a e l e 0 0 0 0 0 0 0 0 0 0
b  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  H e l l o , _ w o r l d 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

(Empty rows g-z ommitted for clarity)

To explain what you see: The first four rows of the disk are given over for a "Table of contents", or TOC. This TOC stores the location of every file on the filesystem. In the above example, the TOC contains one file, named "hello.txt", and says that the contents of this file are to be found between ae and le. We look at these locations, and see that the file contents are "Hello, world"

So far so good? Now let's add another file:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t a e l e b y e . t x t m e z
b  e 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  H e l l o , _ w o r l d G o o d b y e , _ w o r l d
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

As you can see, the second file has been added immediately after the first one. The idea here is that if all your files are kept together, then accessing them will be quicker and easier: The slowest part of the hard drive is the stylus, the less it has to move, the quicker your read/write times will be.

The problem this causes can be seen when we decide to edit our first file. Let's say we want to add some exclamation marks so our "Hello" seems more enthusiastic. We now have a problem: There's no room for these exclamation marks on our filesystem: The "bye.txt" file is in the way. We now have only two options, neither is ideal:

   1. We delete the file from its original position, and tack the new, bigger file on to the end of the second file - lots of reading and writing involved
   2. We fragment the file, so that it exists in two places but there are no empty spaces - quick to do, but will slow down all subsequent file accesses.

To illustrate: Here is approach one

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t a f n f b y e . t x t m e z
b  e 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  0 0 0 0 0 0 0 0 0 0 0 0 G o o d b y e , _ w o r l d
f  H e l l o , _ w o r l d ! ! 0 0 0 0 0 0 0 0 0 0 0 0

And here is approach two:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t a e l e a f b f b y e . t x
b  t m e z e 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  H e l l o , _ w o r l d G o o d b y e , _ w o r l d
f  ! ! 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Approach two is why such filesystems need defragging regularly. All files are placed right next to each other, so any time a file is enlarged, it fragments. And if a file is reduced, it leaves a gap. Soon the hard drive becomes a mass of fragments and gaps, and performance starts to suffer.

Let's see what happens when we use a different philosophy. The first type of filesystem is ideal if you have a single user, accessing files in more-or-less the order they were created in, one after the other, with very few edits. Linux, however, was always intended as a multi-user system: It was gauranteed that you would have more than one user trying to access more than one file at the same time. So a different approach to storing files is needed. When we create "hello.txt" on a more Linux-focussed filesystem, it looks like this:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t h n s n 0 0 0 0 0 0 0 0 0 0
b  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
n  0 0 0 0 0 0 0 H e l l o , _ w o r l d 0 0 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

And then when another file is added:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t h n s n b y e . t x t d u q
b  u 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
n  0 0 0 0 0 0 0 H e l l o , _ w o r l d 0 0 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 G o o d b y e , _ w o r l d 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

The cleverness of this approach is that the disk's stylus can sit in the middle, and most files, on average, will be fairly nearby: That's how averages work, after all.

Plus when we add our exclamation marks to this filesystem, observe how much trouble it causes:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t h n u n b y e . t x t d u q
b  u 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T O C
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
n  0 0 0 0 0 0 0 H e l l o , _ w o r l d ! ! 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 G o o d b y e , _ w o r l d 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

That's right: Absolutely none.

The first filesystem tries to put all files as close to the start of the hard drive as it can, thus it constantly fragments files when they grow larger and there's no free space available.

The second scatters files all over the disk so there's plenty of free space if the file's size changes. It can also re-arrange files on-the-fly, since it has plenty of empty space to shuffle around. Defragging the first type of filesystem is a more intensive process and not really practical to run during normal use.

Fragmentation thus only becomes an issue on ths latter type of system when a disk is so full that there just aren't any gaps a large file can be put into without splitting it up. So long as the disk is less than about 80% full, this is unlikely to happen.

It is also worth knowing that even when an OS says a drive is completely defragmented, due to the nature of hard drive geometry, fragmentation may still be present: A typical hard drive actually has multiple disks, AKA platters, inside it.

Let's say that our example hard drive is actually on two platters, with aa to zm being the first and an to zz the second:

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
b  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

   a b c d e f g h i j k l m n o p q r s t u v w x y z

n  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

The following file would be considered non-fragmented, because it goes from row m to row n, but this ignores the fact that the stylus will have to move from the very end of the platter to the very beginning in order to read this file.

   a b c d e f g h i j k l m n o p q r s t u v w x y z

a  T O C h e l l o . t x t r m e n 0 0 0 0 0 0 0 0 0 0
b  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
d  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
e  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
f  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
g  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
h  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
i  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
j  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
k  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
l  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
m  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 H e l l o , _ w o

   a b c d e f g h i j k l m n o p q r s t u v w x y z

n  r l d ! ! 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
o  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
p  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
q  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
r  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
s  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
t  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
u  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
v  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
w  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
x  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
y  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
z  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

I hope this has helped you to understand why some filesystems can suffer badly from fragmentation, whilst others barely suffer at all; and why no defragging software came with your Linux installation. If not, I'm always open to suggestions

nntp

You talk like Windows has only FAT32! NTFS has MFT to keep all the small files so they don't get scattered on the HD forcing bigger files to be splitted creating fragmentation.

Also it depends more on the OS than the FS is fragmentation is created (for example the OS could save the files in different places instead of putting them all adiacently). In fact .NET 2.0 CreateFile function asks for the "buffer size" (pratically the size of the file) to have windows start writing it in a place where it wouldn't get fragmented.

And finally Windows Vista should have defrag scheduled so fragmentation will finally be a problem of the past.

nntp

贴了两篇老外的解释，我觉得对于不熟悉linux文件系统或者对文件系统根本不了解的新手来说，比较容易理解。

如果有几年linux经验的同学，建议你买一本Robert Lover的专著 Linux kernel development ,国内有英文原版的，你看文件系统那个章节，解释的很清楚。

如果还有人懒到google都不用而继续井底之蛙在这里不懂装懂，我想你明白一个事情，windows系统从造出来开始，因为是个人用的，所以体系结构有很多先天缺陷，他不是一个multi-user, multi-task的系统，虽然微软投入巨资打造了win server.但是微软的生意注定他要背负以前的历史包袱很久很久(XP/Vista就闭嘴一边呆着去吧). Linux最早就是不满意MINIX系统而造出来的，一出生他的设计体系代码实现就是考虑多用户multi-user, multi-task的，所以在文件系统这部分也充分体现。

这里有平时在AIX, HP-UX, Solaris乃至IRIX工作的朋友么?你们有看到这些Unix上有defrag么?

之前很多朋友的讨论都混淆了一个概念: 碎片原理 和产生碎片是两马事情.一个是纸上谈兵，一个是现实世界中系统的实现。

这么多linux跑在现实世界中的各种重要系统上的，而且很多都是一跑起来就根本很难停，你看到那个重要系统的linux管理员三天两头去defrag了? 在XP和Vista上download电影，办公办公，上上网，就三天两头出现碎片，各位知道么？国内就一个期货公司的报盘系统，跑在linux平台上，每分钟会有3000个200字节以内的文件会刷到硬盘上，如果按照微软的弱智设计，那些报盘机岂不是每隔几个小时就defrag?按照那种微软系统的思维，一天交易下来，岂不是机器I/O慢的像牛一样? 实际上报盘机的管理人员，谁去管linux上的什么碎片，根本不用操心的。

过去10多年中，我就看到很多linux管理员担心怎么加固安全用心思在anti-intrusion上的，没有见过谁天天为什么碎片操心的。

楼上的讨论其实很无聊，讨论这么多理论，自己装一个linux,找一个脚本工具，产生些尺寸大小不一的垃圾文件试一下不是更加直观。

我发现很多新同学还是不太习惯linux的学习方法，linux学习的方法不是纸上谈兵，就是动手，几百次，几千次动手。讲究的是体验，不是理论。