I am going to attach a drawing and offer up a challenge to anyone who wants to solve it. I have two ways on how to make it work.
So here is the scenario.
1) This is a transit network. That means there are objects that circumnavigate a known path along tracks.
2) There is already an infrastructure of fiber connected to Cisco ONSs at OC-48 speeds
3) Hanging off the ONS are 6 Cisco 6509 Chassis
4) Hanging off the 6509 Chassis are Cisco 3560 switches
5) Hanging off the 3560 switches are Cisco Wireless Access Points
6) The objects circumnavigating the tracks have a Cisco Wireless Access Point and a Cisco 3560 switch to connect equipment on board
7) The Wireless access points are managed by a WISM on the 6509s
8) Wireless Antenna are spread along the track so that the vehicle has constant contact with the network
9) There are multiple VLANs in the Primary and Backup Data Centers
10) The is just one VLAN on the vehicle
11) The equipment on the vehicle needs to communicate to the data center and to other vehicles
12) Each vehicle has a need for 16 hosts
The question I have (or the challenge) is to figure out how the vehicles can maintain constant contact with the rest of the network. The data center configuration is more or less done. It challenge is figuring out how to subnet and route and/or bridge the trains to keep in contact with the rest of the network while it switches from one access point to another and from one 6509 backbone switch to another.
I assume all the 6500 switches are connected together, but you do not describe how this logical topology is. Is it a shared layer 2 Ethernet or is there a point to point structure between them?
How many access points do you have behind each of the 3560 switches outside the 6500 switches?
What types of access point are used?
The best way to do this is to implement WISM modules (wireless controllers) in the 6500 switches or in at least two of them (redundancy). How many depends on the amount of access points you have installed that are fixed and not moving around. These access points should then be lightweight access points.
Access points on the moving objects will connect with the most optimal fixed access point and traffic will be tunneled back to the controller where the access point is connected. This is called local mode and you can consider the AP on the moving object as a client that are directly connected (locally) at the controllers interface towards the switch.
If the moving object move so that another fixed access point is used, and that access point is connected to the same controller then the client will appear to the network as it have newer moved since it remains connected to the same controller.
More 'challenging' (but solved) is when the moving object connects to an AP that is associated to another controller than where it first connected. Then this controller tunnels the traffic back to the first controller and the client still seems to be locally connected at the first controller.
Summary so far: The point where the AP on the moving object seems to be connected with the LAN will remain at the Ethernet port of the first controller it is connected via as long as it is connected to any of the fixed access points.
IP addressing: You subnet the network as usual and provide a DHCP pool for the clients at each site with 6500 switches that contains WISM modules.
Each 3650 has a 4 Wireless Access points and they are LWAPP tunneled Cisco 1310s
We have two WiSMs, one at the Primary Data Center and a second at the Secondary Data Center.
I want to keep the one-controller model by using a mobility group.
Personally, I wanted to use a flat layer 2 network for the vehicle communications...I might have 100 vehicles but they only have three IPs needed per vehicle. This way there would be no latency waiting for DHCP. Fixed IPs would do the job just fine.
I will do some more research on local mode. I had come to many of the same conclusions as you. I can't remember if that was the solution or not.
Originally I had looked at mobile IP but the WiSM seems to take care of the mobility aspect of the moving APs.