© 2005 Cisco Systems, Inc. All rights reserved. BGP v BGP Transit Autonomous Systems Forwarding Packets in a Transit AS

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Outline Overview Packet Forwarding in a Transit AS Recursive Lookup in Cisco IOS Software Routing Protocols in a Transit AS BGP and IGP Interaction Problems with BGP and IGP Interaction Summary

© 2005 Cisco Systems, Inc. All rights reserved. BGP v All core routers need external routers for proper packet forwarding. Redistributing can overload IGP resources. IBGP is preferred for scalability. Packet Forwarding in a Transit AS

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Routes learned via BGP do not have an outgoing interface associated with them in the routing table. Recursive lookup is performed to forward IP packets toward external destinations. Packet Forwarding in a Transit AS (Cont.)

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Recursive Lookup in Cisco IOS Software

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Recursive Lookup in Cisco IOS Software (Cont.) Traditional Cisco IOS software switching mechanisms perform recursive lookup when forwarding the first packet. –Fast switching, optimum switching. CEF precomputes the routing table. –All recursive lookups are performed while the routing table is built.

© 2005 Cisco Systems, Inc. All rights reserved. BGP v With IBGP running on all core routers, is an IGP still needed in the core? An IGP is needed to resolve BGP next hops and perform fast convergence after a failure in the core network. Routing Protocols in a Transit AS

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Routing Protocols in a Transit AS (Cont.) Core routers need to run BGP and an IGP. BGP carries all external routes. The IGP propagates BGP next hops and other core subnets only. All customer routes are also carried in BGP. –Reduces IGP topology database –Removes customer-caused route flaps from IGP; IGP becomes more stable

© 2005 Cisco Systems, Inc. All rights reserved. BGP v BGP and IGP Interaction Ideally, there will be no interaction between BGP and the IGP. BGP carries external and customer routes. The IGP carries only core subnets. The IGP is not affected by external route flaps. BGP is not affected by failures internal to the network as long as the BGP next hop remains reachable. The only link between BGP and the IGP should be the recursive lookup.

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Sometimes, BGP and the IGP will propagate the same route. Usually stems from bad network design. In this case, routes are determined in EBGP/IGP/IBGP order based on administrative distances of the routes. BGP and IGP Interaction (Cont.)

© 2005 Cisco Systems, Inc. All rights reserved. BGP v If an IGP route is learned through EBGP, the EBGP route will take precedence. Potential causes include bad network design, routing problems, or denial-of-service attack. Protect IGP routes with inbound prefix-list filters at AS edges. Routers should never accept information about local subnets from an external source. Problems with BGP and IGP Interaction

© 2005 Cisco Systems, Inc. All rights reserved. BGP v Summary All core routers need external routers for proper packet forwarding. A recursive lookup is performed in BGP to resolve the forwarding path reference of the next-hop attribute. Packet forwarding to external destinations benefits from the high-speed convergence offered by an IGP; therefore, an IGP is still needed inside a transit AS. The IGP should provide reachability toward BGP next-hop addresses only if they are not disturbed by external updates from other autonomous systems (those are handled by BGP). IP packets could be erroneously forwarded out of the local AS if an external AS accidentally (or by intent: DoS) feeds the local AS with EBGP routes that should be local.

© 2005 Cisco Systems, Inc. All rights reserved. BGP v