as you know checkpoints are great for single file/directory level restores, customers can do restores themselves which is great. You can setup snapsure schedule in Celerra Manager that will create checkpoints at the time you specify, you also get to specify how long the snapshot is kept until it gets refreshed. For example we have multiple CIFS file systems that we sell to different departments. Our standard is to setup 14-days snapshot schedule, this will allow customers to recover deleted or currupted files or simply obtain an earlier version of the file. They will be able to go 14 days back to do that ..that's our standard but really Celerra allows you to have up to 96 checkpoints per file system.  We have the same standard for file systems that are exported via NFS.

So all this is nice and dandy but what happens if Celerra catches on fire and melts down, your primary data is gone so are your snapshots. That's where NDMP comes in,  you can send backup over FC network or over IP network (3-way NDMP) to your tape device (hopefully at either another location or to a tape class that gets send offsite). Once you get Netbackup going and setup NDMP jobs,  Netbackup will instruct Celerra to create a temporary snapshot, send it to tape and then tell Celerra to get rid of it.

You are correct that checkpoints will use SavVol space, how much ? It depends what's going on with your file system, what users are doing, how long you keep checkpoints. You can restrict the size of SavVol ..how big it can grow.

Unfortunatelly you can not use NDMP to backup iSCSI LUNs, you will need to use Netbackup agent on the host where iSCSI LUN is presented.

sorry if i mentioned things you already knew.