One quick note to start: In reviewing the release notes and config guide for the cisco 1300 series routers, RSTP did not seem to be a valid stp option - only per vlan stp seems to be supported.
I came across this URL:
http://www.cisco.com/univercd/cc/td/doc/product/wireless/airo1300/brscg/o13rf.htm
and I have this question:
Are you going to configure the 1300 devcies as shown in figure 6-3 (the workgroup bridge config) or were you thinking of using them as acess points as shown in figure 6-2? You mentioned using them only if the fiber fails, so I assume you want the workgroup config.
Only the 3550 switches are the layer 3 devics in use, thus only they will be running the routing protocols. You stated that the Nortel devices is layer 2 device; but you also mentioned about replacing it with a 3550.
Your topology should be described like this:
1. The connections between the 1300 and the 3550 will be layer 2 only. Each 1300-to-3550 connection will be on a the same vlan that is seperaterate from the wired end-user vlans on the 3550's. This vlan will exist to process the eigrp route protocol hello packets and topology changes.
2. The end-user vlans on each 3550 will be on seperate vlans and the 3550's will exchange subnet reachability via eigrp across their own set of fiber links as well as the radio links. The 3550 fiber links will be another seperate vlan from end-user and the 1300 vlans and again it will exists for eigrp processing only.
3. If possible combine multiple fiber links into a etherchannel config to be viewed as a single logical port from the stp. In addition configure it as an access trunk where only one vlan is allowed to cross, instead of a multi-vlan trunk as only layer 3 traffic will cross it, not layer 2.
I am not sure if this (the etherchannel using only one vlan) is possible, so if it is not then connect each fiber link as a seperate point-to-point connection on a different subnet.
4. EIGRP will be configured to see all 3550-to-3550 connections (fiber and by way of the 1300) as being in the same eigrp autonomous system to make all routes as internal ones.
How many 3550 switches will you use once you acquire all of them, including the replacement for the Nortel device?
I do not think that recovery times will be that much shorter if you try to load-balance equally across all links (that is, using equal-cost multi-pathing) or if you have only one route installed on the routing table and adjust the variance/metrics to prefer one path over the other. I believe this is due to how EIGRP uses the concept of feasible successorts and stores the info in its topology table. I will see what I can find on this topic and post the results here.
Thus any method that the 3550's use to connect will be at layer 3 not layer 2 - hence you have a layer 3 backbone.
I do have these questions regarding the Nortel box: Will it be replaced very soon or will it have to participate in the topology for a while? Can it process per vlan spanning tree BPDU frames?
I do not believe that you cannot use ip unnumbered for high-speed lan interfaces it only works on serial or other low-speed links. I suggest that you carve out a seperate subnet in the IETF private address range for the switch-to-switch connections.
One additional thought I have with regards to the 1300 devices: You are able to configure subinterfaces of the lan interface. I would suggest that you do so to allow for a management vlan and you can apply acls on the 3550's to only insure that management stations can access the 3550 and 1300 devices for status inquiry and config/code changes and updates.
Let me know if this was of help.
