We run a RS6000 with AIX 3.2.5 and ADSM server 2.1.0.11.
We're going from exabyte to 3590. The 3590 is a single drive
but I have several exabyte drives. To take advantage of the
greater speed of 3590's I use a disk pool to move data in from
the exabytes and migrate out to the 3590.

Now I move data from two exabyte cartridges. This data will
be moved to the disk pool until a file has to be moved that is
larger than MAXSIZE. If one of the two processes has to move
a file larger than MAXSIZE it waits until the 3590 is free
and then moves this file directly to the 3590. That's all
right, but then I see that all succeeding files are copied
directly to the 3590 too regardless of their sizes.

I expected that files smaller than MAXSIZE should again be
moved to the disk pool.

The consequences of the described situation is that the
second process is filling up the diskpool. When HIMIG is
reached, a process is started to migrate data to the 3590.
This process has to wait until the first process ends.
During this time the second process fills the disk pool to
100% and hangs around waiting for the 3590 too. After the
first process ends, the second process gets access to the
3590 and migration has to wait until the data from the
second cartridge is moved directly to 3590 too.
So moving two cartridges isn't as fast as could be and there
often are inactive processes hanging around.

I can handle the situation. I watch my processes and cancel
them if necessary and there's not very much data to move
yet. But I want to inform IBM of this problem and I hope
that in a future release it will be solved.

Comments are welcome.

Kind regards,
  Lothar

-----------------------------------------------------------------
Lothar Butsch
Lothar Butsch
University of the German Federal Armed Forces at Hamburg, Germany
-----------------------------------------------------------------
           email: Lothar.Butsch AT unbw-hamburg DOT de
           email: Lothar.Butsch AT unbw-hamburg DOT de