Path selection

1.5.b i Attributes

General information on “BGP Attributes”:

• If two or more paths exist, different BGP attributes will be compared in an exact order to choose the best path
• BGP attribute types:
  • Well-known mandatory:
    • Must be recognized by all BGP routers.
    • Must be present in all BGP updates.
    • Must be passed on to other BGP routers.
    • Example: AS_PATH, ORIGIN, NEXT_HOP.
  • Well-known discretionary:
    • Must be recognized by all BGP routers.
    • Must NOT be present in all BGP updates.
    • Must be passed on to other BGP routers.
    • Example: LOCAL_PREFERENCE.
  • Optional transitive:
    • Must not be recognized by all BGP routers.
    • Must be passed on to other BGP routers (even if not recognized).
    • Example: AGGREGATOR, COMMUNITY.
  • Optional non-transitive:
    • Must not be recognized by all BGP routers.
    • Must NOT be passed on to other BGP routers.
    • Example: MED, ORIGINATOR_ID.

1.5.b ii Best path selection algorithm

General information on “BGP Best path selection algorithm”:

• Controls which route goes from the BGP table to the routing table and is advertised to other peers
• For a path to be selected as “best”, its next-hop must be in the routing table and reachable via recursive routing
• AS-Path must not contain local-AS (this issue can be overcome with allowas-in or as-override)
• BGP Path selection attributes (in tie-breaker order):
  1. WEIGHT (locally significant, Cisco proprietary, higher is better)
  2. LOCAL PREFERENCE (significant within an AS, not advertised to eBGP, carried through confederation, higher is better)
  3. ORIGINATED (where the route comes from, local is considered best)
  4. AS-PATH (shorter is better, AS numbers make no difference, eg. AS_PATH 10 20 is equal to AS_PATH 1000 2000)
  5. ORIGIN (IGP > EGP > Incomplete)
  6. MED (=metric/distance, therefor lower value will be preferred, only compared between peerings to the same provider/AS, IGNORED by default (can be changed) if two different AS’es advertise different MEDs for a single prefix)
  7. PATHS (eBGP > eBGP Confederations > iBGP)
  8. IGP METRIC (prefer the path with the lowest IGP metric to the BGP NEXT_HOP attribute)
  9. OLDEST PATH (prefer the path that was received first = oldest) (Important: Skipped if bgp bestpath compare-routerid is enabled)
  10. ROUTER-ID (lower is better)
  11. LOWEST NEIGHBOR IP (prefer the path with the lowest neighbor IP address)

MNEMONIC:

We Love Oranges AS Oranges Mean Pure Intelligent Organic Refreshment Levels

1.5.b iii Load balancing

General information on “BGP Load balancing”:

• By default, BGP only installs the best route in the routing table
• Pseudo load-balancing can be done with path manipulation
• Example: A BGP peer has several neighbors advertising different prefixes. Either the neighbors or the peer can modify different values of different prefixes so that for every prefix another link is preferred.
• Real load-balancing can be archived with the multipath or add-path feature
• This is explained under 1.5.f Other BGP features"