One other thought...if you are doing ndmpcopy on the same controller, you are limited by the loopback adapter speed..maybe 30-40Mbyte/sec....but I would still probably use qtree snapmirror from 32 to 64 bit...saves you from having to copy the data twice and you can use incremental updates with snapmirror instead of the limited dump levels 0,1,2 for ndmpcopy....
You would create several mirror relationships..one for each qtree then one for non-qtree data using /vol/sourcevol/- on the source and the target will be a qtree. All non qtree data goes to a qtree. Then the cifs shares and nfs exports to non-qtree data have to go one level deeper on the target in a qtree... if luns then map them to the igroup the same as you did before but inside a qtree now. You could ndmpcopy the data up a level out of the qtree or copy it out if you want the same structure. For qtrees on the source, they mirror 1 to 1 to qtrees on the target.
You could ndmpcopy the data up a level out of the qtree or copy it out if you want the same structure.
It is better move than copy. Copy will require double amount of space, and move is quite cheap and fast (as long as number of top level directory entries is reasonable).
I understand the concept of what you are saying, but I was unaware of a command to move data between volumes and Qtrees. Aside from volcopy, ndmpcopy, snapmirror (VSM and QSM). What command would you use to perform this move?
You could use Qtree SnapMirror (QSM) and throttle that, or see if the network team can setup QoS to limit ndmpcopy on port 10000. But know way I know of either to throttle ndmpcopy on the controller.... other than flexshare but that would not be on the network side.