Leonard Chatagnier wrote:
Thanks for the reply as I would like to get to the bottom of this. I
get the general message. However, contrary to what has been said, I
have at least one partition with 4.5% fragmentation(e2fsck reports)
which I don't consider a trivial amound.
It is a trivial amount. On ext3 filesystems with some fairly regular use
you should see fragmentation always around 5%. This is normal, and has
no impact in performance.
On DOS filesystems, the defrag program puts everything on the start of
the drive. You may think that this is good for performance, but consider
this:
- When a file has to be extended (after the fs having been defragged),
the next block will be put after all the other files. This will is much
worse than having a small amount of fragmented files on an ext3 fs,
which will have all their blocks somewhat near.
This means that 5% fragmentation on an ext3 fs means basically nothing,
whereas it may mean some performance degradation on a DOS fs.
Can
you elaborate on why extension 3 file systems wont become fragmented
over time.
*nix filesystems in general distribute files and their blocks in the fs,
in ways that let them grow over time and reduce fragmentation. This
could mean that some blocks are left free after some files are created,
or that writes are deferred slightly in order to better know where to
allocate new blocks and know if the file has potential to grow a lot or not.
Every fs has its own way to do it, but they don't suffer from
performance degradation from fragmentation. Period.
Actually, there is one situation where this may actually happen: on very
active filesystems, with almost no free space.
In this case, fragmentation becomes an issue, but this mostly _never_
happens in real life, and when it does... dump/restore (move everything
somewhere else, and then move it back).
Do I have to revert to ext 2 fs in order to defrag a hard
drive safely? Debian appears to put out the message that ext3 file
system is the greatest yet I don't seem to be able to use any utilities
to maintain it.
You can't use e2defrag, which hardly qualifies as "any utilities".
Let me state this one more time: e2defrag is merely a proof-of-concept,
nobody actually uses it (which is why it doesn't support the latest
additions to ext3).
I know it is difficult to accept this, coming from a Windows background,
but this is how things work in the unix world (or everywhere besides
MS-land, for that matter).
Or if asking for comments on this is more than you care
to do, perhaps you can offer some references(normal human readable
types) that explains my concerns.
The references on this are mostly highly technical, and you can find a
bunch of those by googling. But sometimes the best way to is just to try
and forget the habits learned on MS-land and adopt a wait-and-see approach.
Just let your filesystems be, and in time you will realise that they
works very well without the rituals you are used to perform on your
Windows systems.
I know what I'm talking about, because I already asked the same
questions you are asking now and it took me quite some time to finally
believe it to be true.
And after 8 years using Linux all the time, I came to find the MS-land
rituals somewhat exotic (if unix filesystems take care of themselves,
why can't the so called New Technology File System?).
Why does Debian put out programs that will trash the system without at
least giving some warning?
If it didn't give any warning, you would have a trashed fs by now.
Refusing to work is enough warning, if you ask me.
Just to finish, I understand you wanting to know more before believing
what we are saying, but understand that this is something that most unix
folks know as fact, proven by everyday experience.
Most people don't actually know the specifics on why this is so, but
know that their many-years-old filesystems don't turn slow just from
using them.
Carlos Rodrigues
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
