During the last years, there has been a need to create a scheme that allows the transport of Multicast traffic together with the Unicast traffic over Layer 3 VPNs. This is due to the fact that all the services that are transported by the providers (Private IP, Internet, Frame Relay, etc.) usually use the same IP/MPLS infrastructure and can be offered using Layer 2 VPNs, Layer 3 VPNs of VPLS.
The draft-Rosen scheme has been used for some time as the mechanism to transport Multicast traffic between L3 VPN sites. This method is not based on MPLS or BGP, therefore it requires a different additional infrastructure. In draft-Rosen, the use of the multicast signaling protocol PIM (Protocol Independent Multicast) is required in the client and in the network of the provider in order to create an end to end traffic path. Additionally, it is required to use Generic Routing Encapsulation (GRE) Multicast tunnels to transport the datagrams.
With the current MPLS forwarding and VPNs schemes, and the possibility to create Point-to-multipoint LSPs, the MPLS networks have the possibility to offer multicast services. Adding these new possibilities to the limitations of scalability and complexity of the draft-Rosen scheme, the need to develop a new mechanism for VPN multicast transport rises.
In Next Generation Multicast VPN (NG-MVPN), the same MPLS and BGP infrastructure that in layer 2 VPN, layer 3 VPN and VPLS is used. NG-MVPN uses MP-BGP as signaling protocol of Multicast trees. For this end, seven types of BGP Network Layer Reachability Information (NLRI) messages were standardized, as extensions for NG-MVPN. These messages have different functions like autodiscovery membership, selective tunnel autodiscovery, conversion of PIM join messages, and announcements of active sources. The MP-BGP sessions that are used among the PE routers of the network are the same that are used to signal the different techniques of VPNs on MPLS, and the same Route Reflectors can be used. In this model, the existing limitations in the previous draft-Rosen schemes are avoided (like PIM adjacencies between PE routers), a scalable topology is accomplished and the same IP/MPLS/BGP infrastructure is used in the network.
More information at: http://tools.ietf.org/html/rfc6513