|"...out of the question due to it's cost"|
minute you mentioned cost your project was over.
Zero downtime is HUGELY expensive
ISDN would be a bad guess. Expensive and low bandwidth.
You would start with internal site failover.
Servers in a cluster with shared storage is the norm here. Redundant hardware. UpS and failover power. Hourly backups that are replicated off site. Hardware vendors/capital available for replacing all systems.
Then you move on to site to site communication. Dual providers and dual paths to each site. Consider site to site vpns via the internet, Frame Relay, or dedicated Point to Point T1's depending on bandwidth requirements.
That covers the physical. We still have a bunch of OSI layers ahead :-)
Databases? Needs to replicated throughout.
DNS? Needs to be replicated throughout.
Running Active Directory? Site planning throughout.
You may want to spend some time in Google and high availabiltiy servers to start. This will give you a much better idea of server config since there is no such thing as a primary and backup servers. There are hot failover or standby server in a local scenerio but it takes different planning if hosting web sites or you want the secondary site to come up transparently to the users.