Websphere Business Integration Server

This is my first post to this group -- hope it's not too stupid.

For CONTINUOUS Availability (CA not HA), we think we would like to
geographically "cluster" our two 520's as follows:  
Our website is the only thing running on one of the 520's - call it the
primary node.  

1) First we partition our other 520 and load the website onto it's 2nd
partition - making this the secondary node.  

2) Then, we move the primary node to our ISP's computer room, about 15
miles away, 
and establish a T1 link between the two machines (and the Internet).

3) Next we setup some sort of replication:  cross site mirroring (XSM)
or use the WebSphere Integration Server (which I know nothing about).
Whichever way we do it, transactions need to flow in BOTH directions
simultaneously (setup as peer-to-peer not master/slave).  (We have very
few database writes since 99.5% of the web traffic is serving up static
pages - just a few credit card and address change transactions).

4) Next our ISP assigns both IP addresses to our website -- 2 "A"
records.

5) Then, web traffic is routed to both machines at DSN's whim (somewhat
random I'm guessing).

Now, my hope is that if one node is down (PTFs, comm failure, backups,
whatever) then, web users won't really know because their browser will
have the other IP address as well and just retry there.  I've attempted
many ways to find out why everyone isn't using this simple solution.
There must be a reason.  

I'd appreciate some feedback from anyone that's knowledgeable in
"geographic clustering", or as CISCO calls it GSLB (geographic server
load balancing).

Especially, too, I'd need to learn more about WebSphere Integration
Server.

Help,
Phil