After jumping through various hoops, I have finally gotten the SNAPDIFF feature 
of TSM 6.1 to work. We have a windows client running 6.2.0.1 backing up a CIFS 
share to TSM AIX server running 6.1.3.4.  The first backup runs very fast, 
backing up just over 1M objects, but backs up 366GB of a volume that windows 
shows as having a total capacity of 400GB with only 239GB used. The backup is 
running from the scheduler specifying INCREMENTAL and the -SNAPDIFF option, 
with DIFFSNAPSHOT defaulted (to CREATE).

The next backup, which should be an incremental backs up 246GB and reports 
Objects Inspected 2,648, Objects Backed Up 763,759.  The volume we are testing 
with has no activity going on so I would have expected no files should have 
been backed up.  Also, on this backup there were 580 objects failed (file not 
found error) which did backup on the first backup and still exist.  I found a 
caveat in a technical support brief that SNAPDIFF doesn't support Unicode 
outside of 7-bit ASCII and that ONTAP UTF8 enabled volumes will not return the 
filename in UTF8, but I see nothing in the suspect filenames to make me think 
that is the problem other than they all have embedded blanks somewhere.

So it appears that about 75% of the files are being backed up even though there 
have been no changes and some not insignificant number of files are not backed 
up after the initial full backup.  Anyone out there using this facility and if 
so have you seen these problems?

Thanks
Sam Sheppard
San Diego Data Processing Corp.
(858)-581-9668