Settlements for problems in large-scale BGP networks

You can use the following methods to facilitate management and improve route distribution efficiency on a large-scale BGP network.

• Route summarization

  Route summarization can reduce the BGP routing table size by advertising summary routes rather than more specific routes.

  The system supports both manual and automatic route summarization. Manual route summarization allows you to determine the attribute of a summary route and whether to advertise more specific routes.

• Route dampening

  Route flapping (a route comes up and disappears in the routing table frequently) causes BGP to send many routing updates. It can consume too many resources and affect other operations.

  In most cases, BGP runs in complex networks where route changes are more frequent. To solve the problem caused by route flapping, you can use BGP route dampening to suppress unstable routes.

  BGP route dampening uses a penalty value to judge the stability of a route. The bigger the value, the less stable the route. Each time a route state changes from reachable to unreachable, or a reachable route's attribute changes, BGP adds a penalty value of 1000 to the route. When the penalty value of the route exceeds the suppress value, the route is suppressed and cannot become the optimal route. When the penalty value reaches the upper limit, no penalty value is added.

  If the suppressed route does not flap, its penalty value gradually decreases to half of the suppress value after a period of time. This period is called "Half-life." When the value decreases to the reusable threshold value, the route is usable again.

  Figure 56: BGP route dampening

  

• Peer group

  You can organize BGP peers with the same attributes into a group to simplify their configurations.

  When a peer joins the peer group, the peer obtains the same configuration as the peer group. If the configuration of the peer group is changed, the configuration of group members is changed.

• Community

  You can apply a community list or an extended community list to a routing policy for route control. For more information, see "BGP path attributes."

• Route reflector

  IBGP peers must be fully meshed to maintain connectivity. If n routers exist in an AS, the number of IBGP connections is n(n-1)/2. If a large number of IBGP peers exist, large amounts of network and CPU resources are consumed to maintain sessions.

  Using route reflectors can solve this issue. In an AS, a router acts as a route reflector, and other routers act as clients connecting to the route reflector. The route reflector forwards routing information received from a client to other clients. In this way, all clients can receive routing information from one another without establishing BGP sessions.

  A router that is neither a route reflector nor a client is a non-client, which, as shown in Figure 57, must establish BGP sessions to the route reflector and other non-clients.

  Figure 57: Network diagram for a route reflector

  

  The route reflector and clients form a cluster. Typically a cluster has one route reflector. The ID of the route reflector is the Cluster_ID. You can configure more than one route reflector in a cluster to improve availability, as shown in Figure 58. The configured route reflectors must have the same Cluster_ID to avoid routing loops.

  Figure 58: Network diagram for route reflectors

  

  When the BGP routers in an AS are fully meshed, route reflection is unnecessary because it consumes more bandwidth resources. You can use commands to disable route reflection instead of modifying network configuration or changing network topology.

  After route reflection is disabled between clients, routes can still be reflected between a client and a non-client.

• Confederation

  Confederation is another method to manage growing IBGP connections in an AS. It splits an AS into multiple sub-ASs. In each sub-AS, IBGP peers are fully meshed. As shown in Figure 59, intra-confederation EBGP connections are established between sub-ASs in AS 200.

  Figure 59: Confederation network diagram

  

  A non-confederation BGP speaker does not need to know sub-ASs in the confederation. It considers the confederation as one AS, and the confederation ID as the AS number. In the above figure, AS 200 is the confederation ID.

  Confederation has a deficiency. When you change an AS into a confederation, you must reconfigure the routers, and the topology will be changed.

  In large-scale BGP networks, you can use both route reflector and confederation.