I doubt there is a direct way to replicate data directly to S3.
I was exploring similar options sometime time back but due to very small storage foot-print of the branch office and serving just few VMs (NO CIFS), so CVO was a overkill, so we decided to continue using AltaVault route. Basically, vaulting to AVA and from AVA to S3. Not exactly direct but meets our requirement (secondary backups).
I think for direct S3, I see two approaches:
1) Tiering (Fabric-pool) : This is tiering cold data to S3, but this is not BACKUP, still a primary data suitable to off load infrequent data from SSDs.
2) Using CVO: Yes, this can be a solution.
I would again use the term snapcenter secondary backups, b'cos Primary is the local restore-points (which is on-disk WAFL snapshots) cannot be called backups technically.
In the SnapCenter GUI, under Policies : When you create or edit the policy, under replication tab - there is a Vault/Mirror, this option will allow you to send this data across.
I guess you want to send it to S3:
S3 is file-object : Solution here would be to - SnapMirror data (SnapCenter on-premise snaps) to a destination volume in "Cloud Volumes ONTAP" which will in-turn do the tiering and offloads that data directly to Amazon S3.
CVO: is a EBS (block storage), however S3 is file-object. Therefore, you will be doing :
On-premise SnapCenter Snaps -------> Mirror -----> CVO (Running another ONTAP Image with flexvol) -------> Vaulting------> to S3.
The biggest advantage of using CVO is that it provides 'DR' solution as well. You have the parallel site (for critical apps) running on CLOUD, so even if the on-premise site has to go down due to whatever reasons, you can switch to ONTAP running on the Cloud (i.e CVO).