3Com Switch 7750 Configuration Guide

BGP Overview 353
■ A BGP speaker advertises the routes obtained from EBGP to all its BGP peers
(including both EBGP and IBGP peers);
■ A BGP speaker does not advertise the routes obtained from IBGP to its IBGP
peers;
■ A BGP speaker advertises the routes obtained from IBGP to its EBGP peers (in
Switch 7750, BGP and IGP does not synchronize with each other);
■ Once a BGP speaker sets up a connection to a new peer, it advertises all its BGP
routes to the new peer.
BGP route selection policies
In the implementation on Switch 7750, BGP adopts the following policies for
route selection:
■ Discard next-hop-unreachable routes;
■ Prefer the routes with the highest local-preference;
■ Prefer the routes initiated from the local router;
■ Prefer the routes across the least ASs (that is, the routes with the shortest
AS-Path);
■ Prefer the routes with the lowest Origin type;
■ Prefer the routes with the lowest MED value;
■ Prefer the routes learned from EBGP;
■ Prefer the routes advertised from the router with the lowest BGP ID.
BGP Peer and Peer
Group
Definition
As described in “BGP Routing Mechanism” on page 352, two BGP speakers
capable of exchanging BGP messages with each other are peers of each other. A
BGP peer group is a set of BGP peers.
Relation between peer and peer group
In the Switch 7750, a BGP peer cannot exist independently; it must belong to a
peer group. Therefore, when you configure a BGP peer, you must first create a
BGP peer group, and then add a peer to the group.
BGP peer groups bring convenience for configuration. Once a peer is added to a
peer group, the peer will inherit the same configuration of the peer group. This
can simplify your configuration in many cases. In addition, adding peers to a peer
group can improve route advertisement efficiency.
When the configuration of a peer group changes, the configuration of group
members also changes in the same way. For some attributes, you can configure
them on a particular member by specifying an IP address; and the attribute
settings you made in this way on a member take precedence over the attribute
settings on the peer group. Note that, the members and the group must have
consistent route update policies, but they can have different entrance policies.