I'm reviewing material for the BSCI exam and a question arises that I thought someone here might be able to answer.
Why is route redistribution so complex? Ie. Why wouldnt you simply connect to ASes together by having the border routers run both routing protocols? Wotn that automatically populate the route table with all the routes?
I understand it means more memory and CPU used up on storing networks and path calculation but is that the only reason?
To consider your example as well, having the border router run both sets of protocols you have extra traffic going between the two routers as well, due to two protocols running. Both would run across the link if you have made them in similar AS's, but the catch in the end would be that it would only use the protocol with the lower metric for the routing anyway, assuming you don't get routing loops from what you've created. That said you would likely ditch one of the two routing protocols, but then would need to reconfigure one set of routers, to maklto make it all 1 protocol.
So that long winded example hopefully showed that yes it's more complex, but if done right, can save you from having to reconfigure a group of routers.