Reduced Router Peering ?

[polm]

I am studying for my BCMSN and I came across this :

http://www.cisco.com/warp/public/cc/so/neso/lnso/cpso/gcnd_wp/gcnd_wp8.gif

"Using the routed gigabit ethernet connections between the backbone switches offers the following advantages :

* Reduced router peering for additional stability and scalability"

Can someone explain to me exactly what "router peering" is , and how "Using the routed gigabit ethernet connections between the backbone switches" offers a reduction in router peering ?

 

[yoda291]

If memory serves, router peering is simply the existence of a peering route. Normally reserved for BGP router, a peering route is the route between 2 seperate BGP devices. nowadays however, the term is so ambiguous, it could mean any route agreement. Essentially tho, figure that by making a routed gigabit connection between your core switches means you do not necessarily require routes between your distribution layer switches and all your core switches.

Example, I have 40 or so distribution layer switches in 20 buildings. I have 5 core switches, a 2 switch server module and a 2 switch edge module. Now, if I did not have routed gigabit connections between my core/backbone switches, I would need a routed connection between every distribution switch, server switch, edge switch to every core switch. (40 +2 +2) x 5 = 220 connections. Every distribution layer switch you added would require another 5 connections(10 if you have redundancy). if you had a routed connection between the backbone switches, you could get away with 1 connection per distribution layer switch(provided they were in redundant pairs) and 1 connection per edge and server switch(also assuming redundant pairs). Every switch you add in any module(core,server, or edge) only needs 1 connection to the nearest core switch. 220 connections vs 44 connections. you lose the fast convergence though. You gain orders of magnitude in reliability though.

 

[spidey07]

router peering simply means a neighbor relationship between two routers. that is they are running a routing protocol, monitoring each others reachability and exchanging routes.

A lot of times people will do all sorts of complicated designs where a router is peering with 10 other routers through their redundant links. this is really bad design and leads to routing instability, increased convergence times, routing complexity, etc.

-edit- and its referrign to a layer3 design where core/distribution switches are connected with layer3 links. If they were layer2 links all the core/distribution switches would want to peer with everyone of the others resulting in n(n-1) peers. Bad.