>We are trying to optimize some of our backups and wanted to experiment with
>backuping up JFS filesystems using backup image.  Our thought process was
>that by bypassing the overhead of the JFS we could decrease the backup
>window.  However we are told that one of the rules for TSM (ver 3.7) is that
>if the filesystem is not configured as raw at the LVM level backup image
>will not work.  Is this correct?  Is there away around this?
>How much overhead is actually affiliated with the JFS?

Peter - I presume the context is AIX.  Here, file systems are created within
        Logical Volumes, which is the "raw" space underlying the file system.
The "filespec" used in the Backup Image command is either the file system
name (to implicitly identify the underlying Logical Volume), or the Logical
Volume itself.  If a file system is involved, TSM will proceed to unmount it
so as to back up the Logical Volume unto itself.
See the TSM 3.7 Technical Guide redbook, which as the introduction document
best describes the facility.

Of course, backing up with volume images means restoring volume images.
This is an exception to the TSM philosophy, for when volume images are
more appropriate or expedient, and where file selectivity is not important.
Sure, there is overhead with JFS, as with any file system on top of a
disk partition, but it's value added.

As always, approach Backup on the basis of how you would need to Restore:
do not evaluate Backup methods independently.

   Richard Sims, BU