As far as I know ADSM writes to tape when no stgp volume (not stgpool!) has
enough space to take the backup object which is sent to adsm server. I
don't know how buta works, but - when the object sent by buta is larger
than the remaining free space of the volume or the maxsize parm of the disk
pool, it goes up the storage hierarchy.

If You are using a 4 GB pool for backing up 35 GB data of a single node,
Your stgpool might reach hi threshold soon. As far as I noticed, the
following thing will occur: Migration will start for this node (migration
moves all data of the largest node in the stgp). The migration stops when
all data is migrated, this is after the backup ends when the process moves
data of a currently working node. When the disk volumes become full because
the migration process is too slow, the backup continues to tape. If you
have as much (or more) backup sessions as you have drives, the migration
process will loose its tape drive until the backup process doesn't need as
much drives as you have. So backup continues to tape, migration is waiting.

This is not too easy when one is using small disk pools compared to the
amount of daily backed up data as I do.

You can check if You are backing up to tapes with "q sess f=d" in line
"media access status".

Regards
Roger