In a typical failover, the standby server takes over the IP address from the failed primary server. Routers on the subnet learn the media access control (MAC) address of the NIC that now owns the IP address, and network packets now flow to the new active server. However, failover isn't so easy when the new server is on a different subnet.
One solution is to update your DNS servers so that the host name users typically connect to resolves to the new server's address. This approach can be feasible when all users reside on a private network and the propagation delay for DNS updates is minimal. However, when your users are connected over the Internet, they can wait hours or longer before the host name resolves to the new address, and decreasing that time by setting a short Time to Live (TTL) for the host name increases traffic to your DNS servers.
As another alternative, you can use Virtual LAN (VLAN) technology to route application requests to the remote site. This approach lets servers at the remote site operate as though they're on the local subnet and use its IP addresses.
Other alternatives depend on your router's capabilities. Computer Associates' (CA's) BrightStor High Availability explicitly supports two methods to fail over server IP addresses: Local Area Mobility (LAM) and Floating Subnet. LAM relies on the capabilities of certain Cisco Systems routers and is implemented primarily in the router configuration. Floating Subnet can potentially work on non-Cisco routers. This method relies on the ability of the router to be reconfigured through Telnet during failover and failback and is typically implemented through the scripted application of updated router configuration files.