All selected VPSes are replicates of the storage host and only changes are synced in the chosen period (20 minutes, for example) to storage.

When SERVERware detects that the host is down or unavailable on the network, it will copy VPSes of that host from the storage host to other available hosts and try to start it there.

SERVERware analyzes memory, disk, and CPU of each host before deciding where it will failover each VPS.
VPSes are distributed transparently to other hosts on the network, if those hosts are available.

If there are no hosts available, or if they are full (overloaded already), SERVERware will try to start failover on the storage and controller host (This last step is a desperate move. In practice, the user should take care not to overload his virtualization system to such limits).