bgp
- Russ's course on OE
- as - set of equipment under one organization's control
- as 1 - 65 535 - 16 bit space
- 64 512 - 65 534 - private as s
- dotted notation are more compact - 65 536 = 1.0; a.b, a x 65 536 + b
- nlri = network layer reachibility information ~ route
- as path is not the traffic path, it just ensures loop freens
- bgp connects networks/as s, not routers; that's why there's no loop prevention in ibgp
- customer/transit routes vs infra(structure) routes
- confed(eration) - deprecated cause too complex and doesn't scale
- rr = route reflector
- hierarchical rr
- cluster list in ibgp ~ as path in ebgp
- rs = route server - reflects routes to clients; place of policy management
- ix[p] = internet exchange [point]
- mrai = minimum route advertisement interval - typically 30 sec
- igp almost always converges faster (rarely equal) than bgp
- bgp best path selection is more of a heuristic than algorithm cause different results depending on the when the routes are pushed into the process
- tcp 179
- fast fallover - if the interface fails, it assumes the bgp session also fails
- afi = address family identifier, safi = subsequent afi
- dampening should not be used in the dfz; it slows convergence
- attribute transitivity - is it transfered over ebgp
- cold potato routing - keep the traffic inside the network as long as possible to have more control or to have the traffic delivered directly to the customer
- hot potato routing - closest exit point
- local policy overrides global
- as path poisoning - prepending else's as
- vfr = valley free routing - not routing transit traffic over edge as
- rpf = reverse path forwarding (strict/loose)
- bgp was initially designed for high trust environment