I’m no storage guy unfortunatly but I normally set my disks up as simple as possible, no fancy raid5 or raid10’s or whatever, usually raidz1 mirror zfs for 2 disks at a time.
This would then be called zpool1
I would then split it up into datasets, something like zpool1/lxd zpool1/vm
Then you can give your lxd instance its own dataset to slice up using something like
Regarding backups, its usually a case of either doing a zfs replication using a tool like syncoid, either locally between zpools or to a remote zfs server, and/or doing a zfs send and pigz/tar to an exported file and then copy to backblaze, gdrive or s3fs for some external long term storage.