Routing Protocol Design and Policy
This domain tests IGP area and hierarchy design, BGP policy and scaling, redistribution control, and convergence engineering.
IGP design
OSPF scales through areas: area 0 as backbone, ABRs between areas, and stub, totally stubby, NSSA or totally NSSA at the edge to suppress external or summary flooding. Know the LSA types - 1 router, 2 network, 3 summary, 4 ASBR summary, 5 external, 7 NSSA external - because summarization and filtering depend on them: inter-area summarization happens at the ABR, external summarization at the ASBR. IS-IS is often chosen in very large service provider cores for its TLV extensibility and cleaner area model.
Convergence
Convergence is a design choice, not a default. BFD for sub-second failure detection on links that do not signal failure, loop-free alternation or TI-LFA for pre-computed repair paths, tuned hellos only where the platform can absorb the CPU, and SPF throttling to prevent churn storms. Always pair fast detection with fast reconvergence or you simply detect failure sooner.
BGP policy and scaling
Attributes in decision order: weight, local preference, locally originated, AS_PATH length, origin, MED, eBGP over iBGP, IGP metric to next hop, oldest route, lowest router ID, shortest cluster list, lowest neighbor address. Scale iBGP with route reflectors or confederations; use communities to tag intent (customer, peer, transit) so policy is expressed once. Always apply prefix filters and maximum-prefix limits.
Redistribution
Prefer one-way redistribution. If mutual redistribution is unavoidable across two boundary routers, tag routes on entry and deny the tag on the return direction, set explicit seed metrics, and verify administrative distance interactions - EIGRP internal at 90 will beat OSPF at 110, silently steering traffic.
Worked example
An enterprise migrates from EIGRP to OSPF with mutual redistribution at two core routers. Redistribute EIGRP into OSPF with a route tag of 100 and a seed metric; when redistributing OSPF back into EIGRP, deny tag 100 so OSPF-learned routes cannot feed back. Redistribute OSPF into EIGRP with a tag of 200 and deny that tag in the reverse direction. Without the tags, each router learns the other's redistributed copy, AD comparisons flip, and routes oscillate - the classic symptom is intermittent suboptimal paths and CPU spikes from repeated SPF runs.
Traps that cost marks
- Mutual redistribution without tags or filters, producing feedback loops and flapping.
- Expecting MED to control inbound traffic in another autonomous system.
- Enabling sub-second hellos without BFD or SPF throttling, causing adjacency flaps under load.
- Summarizing in the wrong place - at an ASBR for inter-area routes, or at an ABR for externals.