Hi Tammy
There are 2 options for backup:
VMBackup provides kopies of the VHD that you can manually restore to a new VM or just recover the disks to a storage account.
Recovery Services for you azure VM's are now in preview, this is excellent if you have business critical needs, that in the event MS would loose a Datacenter, you could have replicated VM's and a recovery plan in-place to be up and running much faster and then still have to option to fail back once the primary datacenter returns.
As for you fail over question, you need to use a Loadbalancer.
This load balancer will have its own ip and use's tcp or http probes to detect node health,
the benefit about http, is you can use Custom health probes that return CPU usage, Memory or what ever you can write your way out of.
Edit: Also consider using automation to reduce costs if you dont wish to have both node running at the same time, if you need more VM's overtime i would also consider VM Scale sets, but this requires a bit more technical configuration, the VM's need to be Generelized and configured automatically apon start up.
