Hello Bernard,

you can use at least two mechanisms to improve restore performance
dramatically.

The first is that you run your backups against so called collocated storage
pools - this means that each server will have it's own set of tapes from
which the restore will be performed. At looks very much like having a full
backup set of a server on one or more tapes and avoids defragmentation on
incremental backups that you normally do with TSM. TSM will automatically
take care for this after you have set this definition in the TSM server.

You have to clearly plan which of your servers need to have this
collocation behaviour because this is the most expensive way to keep the
data in TSM. Other clients that do not need this collocation feature should
run on a non collocated pool that allows using one tape for several
machines.

Take a second option:
TSM has a built in feature that is called the backup set. Very easy - this
feature allows you to generate one (or more) tapes with all the data needed
to restore a server on a sequential media. It is easy to extract this data
from the TSM server to a portable media like portable tape or even CD ROM.
You can also put it on a file server as flat file which would be the most
expensive one.
You can now use this portable media and attach it to the server that has to
be restored and do the job without any LAN usage. And with all data
collected to as less media as possible.
(the generation of this backupset can be scheduled in the TSM server)

So these are two option to reduce restore times.



Mit freundlichen Grüßen / best regards
Michael Pirker