There's a lot of questions here. How are you doing back ups now? To tape? To another local disk? Offsite?
ZFS works best (depending on your interpretation of "best") with mirrored pairs (i.e., striped mirrors, aka RAID 10), adding mirrored pairs to your zpool whenever you need more space. RAIDZ has some severe performance problems under some workloads.
If your back up strategy is just to copy to another local disk then just go with ZFS mirroring instead.
If your back up strategy is something else, do a zfs send
after each snapshot.
If you're going to use Solaris 11 or OpenIndiana look into the zfs/auto-snapshot
services (the feature that backends Time Slider). The auto-snapshot
services support running an arbitrary command after each snap (thereby automating sending to tape or remote).
Why is RAID not a backup? on [sf] is applicable. ZFS is not a panacea. Snapshots solve some of the situations where backups can help, but not all. But if you are happy with potentially losing a week's worth of data, then sure, what you propose may be good enough. It's all about time to recovery and to what point in time (in other words, how recent state you can restore to, and how quickly you can get there). What happens if the ZFS pool bombs out for some reason (lightning strike, say) while the weekly backup job is running? – a CVn – 2016-11-29T13:09:38.387