Bill Moran wmoran at
Thu Aug 28 10:49:10 UTC 2008

RW <fbsd06 at> wrote:
> On Wed, 27 Aug 2008 22:08:47 -0400
> Mike Jeays <mike.jeays at> wrote:
> > That's true about FAT.  What I have never understood is why Microsoft
> > didn't fix the problem when they designed NTFS.  UFS and EXT2 both
> > existed at that time, and neither needs periodic defragmentation.
> I think they probably did, NTFS took a lot from UNIX filesystems, and
> at the time it was released they said that NTFS didn't need any
> defragmentation at all. 
> I suspect that it's mostly a matter of attitude. Windows users have an
> irrational obsessive-compulsive attitude to fragmentation, so they
> end-up with good reliable defragmenters, and so less reason not to use
> them. We don't really care, so we end-up with no, or poor,
> defragmenters, which reinforces our don't care attitude.

Companies like Executive software make money off Diskkeeper by running
tests that demonstrate that defragging is wortwhile.  I've seen all sorts
of benchmarks, and I've seen the ones where they demonstrate that NTFS
really does have a fragmentation problem, unlike MS early claims.  If
UFS or ext2 had fragmentation problems, some company would have jumped
on it by now and be marketing a disk defragmenter for Linux/etc.  Also,
accessing a FS by block device and relocating file data isn't a terribly
difficult thing to do, so I have a hard time believing that nobody's
tried it ... my guess is that they've simply found that it wasn't
worth doing.

FAT had a fragmentation problem because it wasn't really designed, it was
just thrown together.  MS was lucky it didn't have bigger problems.

I think NTFS has a fragmentation problem because it wasn't a priority.
MS was focused on building a filesytem that could store the outrageous
ACLs they wanted, and that was non-trival (look at how long it took the
BSDs to have native file-level ACLs).  In that desire to get those other
features in place, I think long-term disk performance got second seat.

Bill Moran

More information about the freebsd-questions mailing list