(posted at 11:50PM GMT) 
We have been using ZFS on our off-site backup platform for the last three months; enjoying such wonderful things as on-disk compression (not used this since Stacker on DOS!), snapshots at filesystem-level (rather than at the block level as LVM does) and more recently, block de-duplication.
There is something very appealing about being able to pull out any version of a file from any point in the history of that filesystem as long as frequent snapshots are being made; my reason for loving it so much is that we provide 5GB of free storage for each ADSL customer and co-lo customer on the basis that if they use it and they need more than 5GB, they are likely to rent space from us rather than go to a different provider of such services.
If a co-lo customer doesn't use it, they end up losing data on their server as a result of either intrusion or physical hardware failure and they have no other backup, we wouldn't have as much sympathy for them as we would for someone who had pulled out all the stops but still got burnt.
You can give a customer free stuff but if they don't use it, you can't force them to, etc, etc.
The primary reason for using ZFS is the snapshot facility; if a customers' server was r00ted and the attacker destroyed what was on the disks and also destroyed the off-site backups (customers have rsync/SSH and SFTP access to their backup account via private SSH key; which would have to be on the server in order to access it); we simply roll back to a snapshot prior to the intrusion and the customer can pull their data from it - once we get their new private key, of course ;-) | 
