Via the Hypervisor is less management overhead (deploying and maintaining software) so it scales better. You’ve got less code running in the guest OS so that can be a security benefit.

Another consideration is that at a hypervisor level, you can schedule backups more effectively. If you have 100VMs and back them all up at 6pm for example, your VBR job can process as few or as many at a time as you desire (if your hardware can handle it) if your hypervisor can handle 5 at most, you cap this via concurrent task processing at a proxy or repository level and youre good. If it’s agent backups each agent can’t see what the other agents are doing so they might sit waiting for resources for a long time or all stressing your systems at the same time and causing a bad experience.

My stance would be to do hypervisor level unless you need an agent specific feature such as support for windows server clusters (the cluster being something you’re protecting, not hyperv clusters etc)

Thank you.  That’s a great explanation!

I have to agree with Michael here as hypervisor level is the easiest for management and scale.  Agent to me should be only for those one off cases form something specific.

Agreed...Hypervisor unless there’s a specific requirement for running the agent backup inside of the OS.  Easier to manage, easier to restore, etc.