Differential backups

[Power2All]

What is your suggestion there?
Is your comment still true if you omit the "OS"? It's very easy to get huge systems...

I think my suggestion is in there already...
A huge system is based on the extra data, not on the OS disk.
For my data, I use either a NFS 4.1 connection, or a direct bind mount, with a eSATA connection from a external device.
VM disks should not be big either way.

 

[LnxBil]

For my data, I use either a NFS 4.1 connection, or a direct bind mount, with a eSATA connection from a external device.
VM disks should not be big either way.

Okay, NFS 4.1 is the only way (out of your examples) that will work in a clustered setup and is a viable option indeed.

I would however run all my applications on a PaaS if you already have HA NFS available and replace the IaaS layer with a Docker-based OS so that the only backup you need to have is your NFS data. Assuming you have a fully automated IaaS setup to replace failed nodes.