With virtual node name you would have to remember who owned what...
under which name specifics are stored... potential for a screw up in a
pressure recovery situation! Plus each additional node name would
suck up a license for a client node.
What might be easier is to split your diskpool into 4 seperate ones.
Take your management class(s) and create 4 (say by adding a 1,2,3 or 4
on the end) then each mgmtclass would point to a specific diskpool.
Each diskpool would have a migration process of 1 but then if data
were in each diskpool you would have 4 migration processes running at
the same time.
On the client you would have to split the data and bind it to the
different management classes by using your include statements. By
filespace sounds like a good level at which to seperate the data...
Now that I've said all that let me add that on our NOVELL servers we
have noticed that it is faster to run a single incremental against the
client than to run multiple, concurrent incrementals (say by logical
volume) in some cases twice as fast 'cause with the multiple,
concurrent incrementals the server just beat itself to death dealing
with lock contention.
You might try just letting a single task run against your client...
Now with archives it is a different story! By far you can get more
data moved with multiple, concurrent sessions but there isn't the mass
adsm db overhead with archives that there is with incrementals.
hope this helps,
later,
Dwight
______________________________ Reply Separator _________________________________
Subject: Can a Migration Process use more than 1 drive?
Author: mformato (mformato AT US.IBM DOT COM) at unix,mime
Date: 12/15/98 3:02 PM
We are backing up to disk pool on an SP2 AIX verson 4.2.1.0 with ADSM
client 3.1.20.3 and server 3.1.1.1on the same machine. Migration to a 3494
tape library uses only one tape drive, eventhough there are 4 available and
the mount limit is set to 4. I understand this is because there is only
one client node registered to this server. Is there a way around this? We
would like the migration process to use all 4 drives. The client backs up
via several sessions based on filespace, so we are thinking of using
virtual nodename option in the command line to give adsm the illusion of
many nodes and therefore the ability to run several migrations. We have
set the number of migration processes to 4.
Does anyone have any other suggestions or similar experience? Thank you.
Marci Formato
|