There are two different things: - live migration: its wonderful because I can update servers firmware without any VMs downtime - VM restart if a host dies - works wonders in a cluster where given enough host capacity you have a level of assurance that the downtime is very brief.

And regarding Fault Tolerance, it is tricky and if you don't need that level of availability it's better to just plan for a VM restart or have other types of high-availability in place to mitigate a host failure.

> You can't have RAM sync between hosts.

That's what FT is for, if the primary VM instance dies or becomes unresponsive, you failover to the mirror instance.