I would assume the goal here is to speed the restoration process. The only
way to guarantee that happens is to ensure you don't have tape contention
during the restore. At the directory level that is hard unless you use
multiple management classes and tape storage pools.
This points out the larger problem in fine detail. If you want to do
multiple simultaneous restores you must plan for them during backup. As
manufacturers introduce 72GB drives, we're going to see enormous RAID sets
(we are already seeing enormous sets but this will make them enormouser -
new word my prerogative). You should/must break these large sets into
multiple filespaces and then use collocate=filespace on your storage pools.
I would guess the best restore times you are likely to see will be around
10GB/hour/stream. If you have a 360GB single filespace RAID set you can
expect a restore time ~36 hours. You are being foolish to expect faster.
The tape drive spec types will crawl out the woodwork and tell us that the
drives can restore much more than that per hour. Sure, but unfortunately
we're dealing with reality here and not the spec sheets. There are many
other factors influencing how fast we move data.
I know, I know, breaking the big drive into smaller ones means more
management. As long as you don't ever need to restore it creating a mongo
drive is easiest thing to do. Just remember you did this because you never
were going to restore it. Try to remember that in the middle of the second
day while you're trying to restore it.
Let's bring some rationality back to our business. Bigger isn't always
better.
Kelly J. Lipp
Storage Solutions Specialists, Inc.
PO Box 51313
Colorado Springs CO 80949-1313
(719) 531-5926
Fax: (240) 539-7175
Email: [EMAIL PROTECTED] or [EMAIL PROTECTED]
www.storsol.com
www.storserver.com
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of
Michael Bartl
Sent: Thursday, September 13, 2001 9:27 AM
To: [EMAIL PROTECTED]
Subject: Re: Restoration of multiple directories through command line?
Nicholas,
I think this method doesn't help.
I came across the same problem some time ago and didn't find a solution,
too.
The problem is:
When using the GUI interface you can select any set of files and folders
to be restored in ONE SESSION.
When using the CL interface you only can select one tree to be restored.
Now think of a situation where the files of your node are spread over 20
tapes. The first restore mounts 15 tapes, the second and third restore
will mount some of those 15 tapes AGAIN for sure, so you loose a lot of
time in comparison to the GUI method.
Best regards,
Michael Bartl
Nicholas Cassimatis wrote:
>
> When I wanted to do this, I wrote a script to run the restore of each
> directory/filesystem, but I had to do it one at a time. The following
> would be my script with your filesystems on AIX:
>
> dsmc restore /oracle/stage/ -subdir=yes -inactive -pitd=09/02/01
> read NOTHING
> dsmc restore /sapmnt/SA2/ -subdir=yes -inactive -pitd=09/02/01
> read NOTHING
> dsmc restore /oracle/SA2/sapreorg/ -subdir=yes -inactive -pitd=09/02/01
> read NOTHING
>
> The "read NOTHING" lines causes the script to pause for input (any
> keystroke) so I was able to verify the restore went without problems (on
NT
> it would be "pause").
>
> Nick Cassimatis
> [EMAIL PROTECTED]
>
> Today is the tomorrow of yesterday.
--
Michael Bartl
Office of Technology, IT Germany/Austria
Cable & Wireless Deutschland GmbH.
Landsberger Str. 155 Tel.: +49 89 92699-806
80687 Muenchen Fax.: +49 89 92699-302
mailto:[EMAIL PROTECTED] http://www.cw.com/de
