WAN Replicated Service Availability?

[Jeff With The Big Yellow Suit]

Two approaches immediately pop to mind.

1)  Routing magic.  If you have control of BGP from the two locations
then you could advertise the addresses of the service providers from
both locations.   The service is provided by machines on network P.
Normally router A advertises itself as the path to P.   When A goes down
B starts advertising itself as the path to P.   In both cases the 
routers sit
in front of a network P.

The satanic details are in how you determine when a transition should
happen.  You could do this by running a routing protocol between the
two border routers and doing much magic.  Another approach is writing
an application which monitors the connection and triggers the switching.
(You can use RIP with extremely limited parameters as a transit medium
between the monitoring host and the routers... or just log into the 
routers
and make the change.   Junipers are nice for the latter.)

2) Application magic.  If you have control of the application source, 
then
have the application attempt to connect to service provider P1, P2, etc.
This could be set in a configuration file or by querying DNS for SRV
records.   (One of the things the world needs is a simple library to 
allow
server location and connection via SRV records.  Does anyone know
if such a beast exists?)

-jeff