You can start multiple restores. The problem is how. If you have multiple
drives/filesystems/mount points, great, start a restore for each one. If
you only have a single, you've got a trickier problem. For single drives,
you would need to restore by directory tree perhaps. Easy if you only have
a couple of trees, hard if you have lots of directories at the top level.
Then the problem becomes one of tape contention. If more than one restore
stream needs a particular volume, you've got contention and something will
have to wait.
The capability to restore to disk first and then to the client doesn't
really exist. If you know which tape volumes have the data you need (and
there isn't a real good way to find only active data), you could do move
data commands on each volume to the disk storage pool. Then start the
restore. The problem here is you probably have other client's data on those
volumes so you might not have enough disk pool space.
You could also generate backupset. See help for how this works. You can
then take the media you generate directly to the client and restore
backupset. TSM still has to mount all of the tapes and write the data. I'm
thinking this isn't going to be much faster than a network based restore.
The moral of this story is to have multiple drives/filesystems/mount points
and use filespace collocation on the tape storage pool. You can then start
multiple restores without tape contention. Most of us really like these
mongo RAID5 disks but they cause real problems during restores. Break them
into smaller chunks so TSM creates more than one filespace.
That or buck up and endure it!
Kelly J. Lipp
Storage Solutions Specialists, Inc.
PO Box 51313
Colorado Springs CO 80949-1313
(719) 531-5926
Fax: (719) 260-5991
Email: [EMAIL PROTECTED]
www.storsol.com
www.storserver.com
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of
Gill, Geoffrey L.
Sent: Monday, August 14, 2000 12:30 PM
To: [EMAIL PROTECTED]
Subject: Restores
If a restore has 30GB isn't it possible to open numerous sessions from a
single node to the ADSM server so the restore goes much faster? Would this
be automatic within ADSM to start these sessions? What are those options and
where are they set, I can't seem to find it?
I'm also wondering what areas I would make changes if I wanted to stage a
restore to a disk pool before it went to the client node. If I know a node
with 30GB of data is going to have some disks replaced isn't it much faster
to restore the data to a disk storage pool first?
I have ADSM 3.1.2.55 on AIX 4.3.3 with SSA's, 3494Lib with 4 3590-E1A's.
Thanks,
Geoff
