John

Zlatko is right. Since only TSM1 knows what is on the virtual volume, only
TSM1 can perform reclamation of the virtual volume.

The TSM designers then had a design choice about how this is acomplished.
The obvious way is for reclamation of the virtual volume to be accomplished
by mounting the virtual volume to be reclaimed on TSM1 as input, along with
a scratch virtual volume for the output. This involves two physical tape
mounts on TSM2, but crucially the data would have to pass over the network
from TSM2 to TSM1 when it was read and then from TSM1 to TSM2 when it was
written.

The way it is actually implemented avoids this duplication. TSM uses the
fact that the virtual volume is a copy storage pool volume and therefore
all the data containted within it will also typically be located on a
primary storage pool volume held locally on TSM1. By using this data
instead of the data held on the virtual volume, only the writing of the
data is performed over the network.

So the process has actually been optimised to reduce network utilisation.

However, it should be remembered that reclamation of the virtual volume
alone is useless without also separately performing reclamtion of  the
physical volumes on TSM2 in the way that you describe.

Going slightly off-topic, it is debatable whether using the primary storage
pool volumes as source for the reclamation is optimal in most real-world
scenarios. Since the primary storage pool (physical) volumes will typically
be co-located and the copy storage pool (virtual) volumes typically won't
be, the 50% reduction in network utilisation is more than negated by the
increased number of physical tape mounts required. That is to say the mount
of a virtual volume as source typically requires only one physical tape
mount on the destination server, whereas to mount the physical volumes
required from the primary storage pool requires as many tape mounts as
there are nodes whose data is contained in the virtual volume (assuming
co-location at the node level).



Having said that, if there are a large number of virtual volumes to be
reclaimed, TSM will optimise the tape mounts from the primary storage pool
to ensure each physical volume is only mounted once.



With all that said, I've spent five years struggling to get my head round
TSM server-to-server comms and now we're just about to rip it out in favour
of SAN-based off-siting of data with library-sharing based on Gresham EDT
and ACSLS!

Neil Schofield
Yorkshire Water Services Ltd.




Visit our web site to find out about our award winning
Cool Schools community campaign at
http://www.yorkshirewater.com/yorkshirewater/schools.html
The information in this e-mail is confidential and may also be legally
privileged. The contents are intended for recipient only and are subject
to the legal notice available at http://www.keldagroup.com/email.htm
Yorkshire Water Services Limited
Registered Office Western House Halifax Road Bradford BD6 2SZ
Registered in England and Wales No 2366682