Some quick answers...
xfs_db can tell you overall fragmentation, I suppose a wrapper around
this would be nice:
[root@stout root]# xfs_db -r /dev/hda9
xfs_db: frag
actual 408, ideal 408, fragmentation factor 0.00%
actual/ideal are number of extents, factor is something like
(actual - ideal) / ideal
xfs_bmap can show you the actual number of extents in a file.
xfs does try to place files in the same directory in the same
"allocation group," which is a sub-chunk of the filesystem.
fsr does not defrag freespace.
xfs doesn't generally need much defragmentation, although certain types
of writes (sync, for example) -can- fragment files pretty badly.
If you'd like to do some of the performance measuerements you talk
about, let us know. :)
-Eric
On Wed, 2002-11-13 at 16:02, LA Walsh wrote:
> Is there any tool to tell how many files have how many fragments out of
> how many files...something on the order of
>
> the following files have more than 1 fragment:
> frags file
> 2 /tmp/xyzzy
> 35 /var/mail/box
> ...
> 37 objects out of 1000 total, >1 fragment: 3.7% fragmented objects
> 1100 fragments total, 10% fragmented file space (1000 optimal
> fragments)
> [are xfs allocates are done in "zones"? so similar files in same
> directory
> are usually near each other on disk?] Does it make sense to talk about
> 'free space fragmentation'?
> For example,
> on a 'start from beginning' FS like FAT32, optimal may be 1 large area
> of free space, except for an extra segment or so after frequently-
> modified-files, but if a disk was, in some way zoned, optimal might
> be 100 areas of freespace]
> 110 free space fragments (50 optimal) 55% fragmention.
>
> Does xfs_fsr defrag freespace as well? Any idea on how much
> fragmentation
> affects performance on xfs? Is it on the same order that it is on FAT32
> or
> NTFS? I've read (perhaps it was defrat company propaganda), that while
> FAT
> could be defragemented on an occasional, as needed basis, NTFS needed
> more
> aggressive fragmentation -- so much so that an auto defragger in
> background
> could be useful in some circumstances.
>
> I've tended to think of *nix fs's as not usually needing defragmenting
> if
> they were kept below 90% capacity, I think xfs's defragmenter is the
> first I've
> heard of on a *nix.
>
> Order on disk correlated to execution order after boot can affect boot
> performance on WindowsXP by over 100%. Have there been any measurements
> on block ordering with xfs (or any *nix fs's for that matter)...
>
> -linda
>
>
>
> > -----Original Message-----
> > From: linux-xfs-bounce@xxxxxxxxxxx
> > [mailto:linux-xfs-bounce@xxxxxxxxxxx] On Behalf Of Chris Wedgwood
> > Sent: Saturday, November 02, 2002 3:04 PM
> > To: mgiesbre
> > Cc: linux-xfs@xxxxxxxxxxx
> > Subject: Re: Defrag Utility
> >
> >
> > On Thu, Nov 02, 2000 at 05:06:02PM -0600, mgiesbre wrote:
> >
> > > I was wondering if there was any work being done on an XFS defrag
> > > utility, like the command that is available in IRIX. Any info?
> >
> > man xfs_fsr
> >
> >
> > --cw
> >
> >
> >
>
--
Eric Sandeen XFS for Linux http://oss.sgi.com/projects/xfs
sandeen@xxxxxxx SGI, Inc. 651-683-3102
|