We are just getting into larger NT machines as you described and are also
looking for "tricks" to improve restore performance.  I'm a NT novice at
best, but its seems as though restore performance on a very large logical
volumes
could be improved if the ADSM client could use multiple threads and/or
processes to restore multiple parts of the logical
volume at the same time.  There has been a lot of discussion on this list
about homegrown or semi-supported methods of
running multiple concurrent processes to improve elapsed time of
backups/restores.  Testing with my own laptop running
Windows 95 I've cut my backup time nearly in half by running a few
concurrent processes.  It seems as though the single
thread process used by the standard methods provided with the product don't
ever max out your network interface card or
system resources therefore, some ADSM users are able to realize better
elapsed times by running multiple concurrent processes.  Maybe the best
long term solution would be for IBM to enhance the ADSM client to use
multiple threads/processes to improve elapsed times.  In the interim we are
attempting to develop an automated homegrown solution for running multiple
client processes to reduce restore elapsed time.

Jeff Connor
Niagara Mohawk Power Corp
Syracuse, NY





Taylor Mitchell <taylor_mitchell AT MSN DOT COM> on 04/22/98 10:10:19 PM

Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>

To:   ADSM-L AT VM.MARIST DOT EDU
cc:    (bcc: Jeffrey P Connor/IT/NMPC)
Subject:  Large ADSM Restores




While much of the discussion in this forum has focused on backup
performance
issues, another issue that is perhaps even more important is restore
performance.  In particular, as servers have become more powerful, have
become consolidated, and larger amounts of disk media are attached, the
implications of a media failure become more serious.  For example,
depending
on the average file size, the cpu speed of the client, and the network
connectivity, the time to restore a larger(e.g. 24 GB) RAID storage device
could  be several hours or even more than one day.  If you have never gone
through the exercise, I would encourage you to restore the contents of a
entire raid storage unit to a test server.  You may be surprised how long
this can take, especially if there are lots of small files.

I'd be interested in hearing from on others on the methods others have used
to mitigate the risk of an extended outage due to long restore times that
could be encountered if an entire raid storage unit failed.

For example, limit the amount of storage that can be attached to a single
server?

Create prioritized plans to restore "important" data first?

Limit use of ADSM to clients below a certain size?

Thanks in advance,

Taylor