FW: Filespace collocation not working as it should

Got a response from IBM re our restore problems on HP40.  Basically it
seems there is an algorithm that gives each restore session a slice of
time before moving to the next process.  I think the philosiphy would
be that they could be sessions from many different clients and
therefore all deserve a fair go.  I think this is wrong and the
sessions should finish once they are started, as only in that way will
they all finish quicker.  Better to sacrifice the one for the many.

The answer is to run 2 concurrent sessions only, as we only have 2 tape
drives.  This can be done using 2 scripts (which run concurrently)
which run the restore of each file system sequentally, rather than
putting them all in the background and running concurrently.  Perhaps
one more test is in order.  This is a learning process for us all!

Regards,
Simon


Response From IBM:


From the Questions,


     For 11 sessions started, each belongs to different tape. ADSM would
select the session desired based on its internal alogrithm.
     Thus you would see ADSM goes from one session to another causing a lot
of mounting and umounting. Activating 11 session would
     not improve the performance but decrease unless the filesystem to be
restored is in cache or you have enough drive to substain.
     Else you have to handle two session by two session at a time by
controlling the restore process.

     Recommedation:
          1)   Keep all the server's backup to 1 single tape. not 1
filesystem to 1 tape. Restore Sequentially.
          2)   Define two ADSM nodes in the same server, some filesystems
belong backup using one node.
               some filesystems backup using the other.
          3)   Restore by two session at a time.


     Thanks and please get back to me if you need further information.

Original question:

Our environment:
============

Server:
ADSM  V3.1.1.5
IBM R40
3494 Tape Library with 2*3590 drives

Client:
HPUX
ADSM V3.1.0.0

We have set up filespace collocation on one of our storage pools. This
storage pool only has one client with about 11 filesystems on it. There
are about 20 tapes available for this filesystem and each filesystem is
less than 1 GB. This means that each filesystem occupies one tape.

We attempted a full restore for this client by running multiple restore
sessions (11 sessions).  Since we only have two tape drives, this means
that the server can only handle two restores at one time while the
other sessions remain in "Media Wait" states.

What we found was that ADSM keeps switching between these 11 sessions.
It does two restore sessions, stops,  does a restore for another
session, stops, goes to another, comes back to the first session and so
on. Furthermore, it seems that it chooses the sessions at random. As
you can expect, the restore generated a LOT of tape mounts as each
filesystem uses an individual tape. This means that it keeps coming
back to the same tape whenever it chooses a session which it has
serviced previously.

My question is why does it not finish off one session first, after all,
as mentioned, each filesystem being restored is on an individual tape.
Is there a parameter that determines the amount of time spent each
session or is this a bug?

Thanks in advance.

Shafiee.
ADSM Support,
Brunei Shell Petroleum.