7.4 Extending BGP: MBGP

7.4 Extending BGP: MBGP MBGP is not a separate protocol but an extension of BGP, so the specifics of MBGP peering are similar to conventional BGP peering. RFC 2858 defines the multiprotocol extensions for BGP4, and the extensions are implemented as optional path attributes. A standard BGP Update message may contain multiple destination prefixes that share the same path attributes such as AS_PATH, NEXT_HOP, MED, and so on. As previously stated, a BGP Update message contains a single instance of each path attribute, plus a list of prefixes that share those particular attribute values. This strategy is unlike that of most IGPs, whose updates contain a list of prefixes, each listed with its own attributes. BGP's method of exchanging updates leads to the efficient use of bandwidth for a protocol with so many attributes, especially considering that many of the attributes are optional and do not pertain to every prefix. MBGP adds two new path attributes called MP_REACH_NLRI and MP_ UNREACH_NLRI. MP_REACH_NLRI is used instead of the standard BGP NLRI for prefixes from protocols other than IPv4 or for IPv4 prefixes intended for a routing table other than the unicast forwarding table. The MP_UNREACH_NLRI attribute is used in place of the Withdrawn Routes field of the standard BGP UPDATE message to indicate that the specified prefixes are unreachable. MBGP can be used to carry forwarding information for any protocol that has a prefix-mask hierarchical address space. Possible protocols include IPX (Novell's Internetwork Packet Exchange) and IPv6. The most popular implementation of MBGP currently is for multicast routing. This application is so popular that it is common to hear MBGP translated to "Multicast Border Gateway Protocol."