Currently we have the time, and our guest partitions are not all that
significant, to do a full system save, including all, on both the guests
and the host. Interesting idea though. I might mull that over.
In the case of catastrophe I would think recovering the guested system as
a storage space would negate the need to to a restore of the guest. And,
would be faster, and more inclusive. Why more inclusive? Well, your BRMS
information, last save date (as updated by the full system save of the
guest), /tmp type stuff used by save/restore, etc would be included. The
only additional stuff would be any guest saves done between the full
system save of the host and the damage to the object(s) you are trying to
And a simple RST of the guested space on the host is MUCH easier than the
step-by-step recovery from the full system save of the guest.
A save of the guest does allow for granularity of restore. For example,
restoring one library to make a test library.
What is the rule of thumb of backup planning? Plan for how you'd complete
your restore. The fastest COMPLETE restore of a guested lpar is by a RST
of the storage space. (unless your host is also hosted, and backed up, by
VIOS [of which I don't know scat]).