All,

I would appreciate hearing/reading other experiences with NDMP TOC
builds and using your TOCs through the TSM Java GUI interface, or
just querying the contents of your TOC.

We're running a dedicated NDMP TSM server 5.3.1.3, pSeries 615, dual
processor, 4GB memory, AIX 5.2,  65GB DB, RAID-5.  All TOCs are
stored on disk.

TOC loads at the GUI take *forever* on large volume restore
queries.  Example, over a 7 day period of level 0 then level 1
backups, all with TOCs:

Full Image -> 641 GB xxxx million files
Diff (1) -> 1.7 GB
Diff (2) -> 3.8 GB
Diff (3) -> 6.7 GB
Diff (4) -> 16.7 GB
Diff (5) -> 19.9 GB
Diff (6) -> 23.4 GB

The load operation at the GUI takes 25 minutes.  That's
unacceptable.  I assume all 7 TOCs need to be looked at and loaded
before displaying all files at the GUI Restore menu.

When I run dsmadmc q toc nasbackup and dump the contents of the
largest TOC to a file on another machine, we observe one CPU on the
TSM server is pegged (kernal time, no other activity on the server),
about 57,000 system calls per second, and about the same number of
bytes dribbling out.  Is this QUERY TOC translating every character
of the TOC with a system call ?

We thought of pre-loading each TOC after backups each day; our TOC
retention is 1440 minutes.  Can anyone say how to load a TOC into the
database outside the GUI?  And if such a capability were
available,  it seems we would need to load every level 1 TOC back to
and including the last level 0 backup?

Thanks for comments.

John