BGP Overview

BGP is a heavy-duty, secure protocol that must be configured on a per-peer basis. Once a peering session has been configured, BGP uses a TCP connection to establish a session. After a BGP session is established, traffic is passed along the BGP-enabled link.

Although BGP requires a full-mesh topology to share route information, you can use route reflectors and confederations in a large autonomous system (AS) to reduce scaling problems.

BGP Peering Sessions

Unlike RIP and OSPF links, BGP peering sessions must be explicitly configured at both ends. To establish a session between BGP peers, you must manually specify the interface address to which you are establishing a connection. Once this configuration is complete on both ends of a link, a TCP negotiation takes place and a BGP session is established.

The type of the BGP peering session depends on whether the peer is outside or inside the host's autonomous system (AS):

• Peering sessions established with hosts outside the local AS are external sessions. Traffic that passes along such links uses external BGP (EBGP) as its protocol.
• Peering sessions established with hosts within the local AS are internal sessions. Traffic that passes along such links uses internal BGP (IBGP) as its protocol.

To monitor BGP neighbors, see the information about real-time performance monitoring (RPM) in the J-series Services Router Administration Guide.

IBGP Full Mesh Requirement

By default, BGP does not readvertise routes that are learned from BGP. To share route information throughout the network, BGP requires a full mesh of internal peering sessions within an AS. To achieve an IBGP full mesh, you configure a direct peering session every host to every other host within the network. These sessions are configured on every router within the network, as type internal.

Route Reflectors and Clusters

In larger networks, the overhead needed to implement the IBGP full-mesh requirement is prohibitive. Many networks use route reflectors to avoid having to configure an internal connection to each node for every new router.

Note: You must have an Advanced BGP Feature license installed on each Services Router that uses a route reflector. For license details, see the J-series Services Router Administration Guide.

A route reflector can readvertise routes learned through BGP to its BGP neighbors. If you define clusters of routers and configure a single router as a route reflector within each cluster, a full mesh is required only between the route reflectors and all their internal peers within the network. The route reflector is responsible for propagating BGP routes throughout the cluster.

For more information about route reflectors, see Route Reflectors—for Added Hierarchy

BGP Confederations

Large ASs can be divided into smaller sub-ASs, which are groups of routers known as confederations. You configure EBGP peering sessions between confederations, and IBGP peering sessions within confederations. Within a confederation, the IBGP full mesh is required. For more information about confederations, see Confederations—for Subdivision