Redundancy
The main purpose of having a second connection to the Internet is to be able to access the Internet from home for browsing etc. should the primary connection – the fiber – for any reason disappear.
If I had hosted my mail and web servers at some web hotel this setup - which does not include any tunnels – would be almost perfect and nothing more would be required.
But I do have servers running at home. I realize that access to the servers is only possible via the fiber since the dongle is NATed – probable several times – so only outgoing connections are possible.
In order to make the switch between connections seamless for the workstations (WS) and servers some means of setting the default route must be provided. Default routes are distributed by DHCP and are hard to chage after bootup. Enter the concept of Virtual Redundant Routers with the supporting protocol VRRP. This gives the workstations and server a fixed default route while VRRP makes sure that there 'always' is an active router to carry the load.
The details of my VRRP implementation may be found in Redundant Routers.
I have my DHCP server on the main router, so the next thing is to make sure that there always is a DHCP server available for the clients (the worksstations etc.). My DHCP server uses ISC DHCP and the ISC implementation of DHCP includes the DHCP Failover Protocol so it is natural to use that one here.
Failover is described on the Failover page.