| Comments

OSPF Path Selection Order of Preference and Summarization - part1

This is part 1 of 2. My shortish summarized version of OSPF Path Selection Order of Preference and Summarization.




Open Shortest Path First (OSPF) Basics:

  • Use protocol number 89
  • Use bandwidth-based cost metric
  • Use highest IP on interfaces in UP/UP state as Router-ID by default
  • Broadcast Networks:
    • 224.0.0.5 – ALL OSPF Routers – Used for Hello packets. Used by DR/BDR for LS-Updates and LS-Acks
    • 224.0.0.6 – ALL DR/BDR Routers – Used by ALL routers except DR/BDR to send LS-Updates and LS-Acks to DR/BDR
  • Point-to-Point Networks:
    • 224.0.0.5 – Used by ALL for ALL OSPF “messages”
  • Non-Broadcast Networks:
    • NO multicast – Destination IP of Hello / Link State packets is unicast IP of a specific neighbor.
    • Neighbor IP is a required part of OSPF configuration for Non-Broadcast links.
  • OSPFv2 Supports cleartext and MD5 authentication
  • Supports MPLS-TE via “Opaque” LSA’s (9-link-local, 10-area-local, 11-AS)
  • If equal cost routes exist, uses CEF load balancing.

Neighbors MUST agree on following to become adjacent:

  • Area Number
  • Authentication
  • Hello and Dead intervals
  • Stub area flag
  • MTU
  • Neighbors MUST use compatible network types
  • MUST be on same subnet. ONLY on Point-to-Point links rule does not apply. (ip unnumbered)

Path Selection:

These rules apply in THIS order even if the OSPF link metric (Cost value) is changed.

  1. Intra Area Routes (O)
  2. Inter Area Routes (O IA)
  3. External Type 1 (E1)
  4. External Type 2 (E2)
  5. NSSA Type 1 (N1)
  6. NSSA Type 2 (N2)

Interface cost is derived from the bandwidth. The formula is:

Cisco default, Reference = 100000 (Kb/s), 100 (Mbps)

Cost = Reference bandwidth / Interface bandwidth. (Rounded down to the closest integer)

Modify reference bandwidth: ospf auto-cost reference-bandwidth

Verify reference bandwidth: show ip ospf | include Reference

Verify interface cost: show ip ospf interface Gi0/0 | include Cost

NOTES:

  • SPF calculations are ONLY performed for Intra-Area routing. ABR advertised information is trusted for Inter-Area and External calculations.
  • Changing cost values, you need to use the ip ospf cost <COST_VALUE> command on the interface and NOT the bandwidth statement. The bandwidth command is also used for other traffic minipulation techniques like QoS and will break those.
  • ONLY point-to-multipoint and point-to-multipoint non-broadcast support OSPF cost value on a per-neighbor basis. (neighbor x.x.x.x cost x)
  • Hello messages can ONLY be suppressed on point-to-point and point-to-multipoint network types. (ip ospf demand-circuit)

On FastE, GigE, TenGigE and 100GigE the default OSPF cost will be “1”, so you WILL want to change this!

Summarization:

All routers inside the area MUST have exactly the same LSDB to be able to summarize!

Summarization can ONLY be done:

  • between areas (ABR) with area <SOURCE_AREA> range <ADDRESS> <MASK>
  • during redistribution (ASBR) from another protocol with summary-address <ADDRESS> <MASK>

It is possible to ‘black hole’ the traffic or sub-optimal route within the domain with OSPF summarization. OSPF will AUTO create a discard route with the Null0 interface as the next-hop to prevent the router from using a shorter match (0.0.0.0), if the more specific destination network is not present in the routing table. This behavior can be disabled with no discard-route <INTERNAL|EXTERNAL>.

Comments