Understanding Multiprotocol BGP-Based Multicast VPNs: Next-Generation
Multiprotocol BGP-based multicast VPNs (also referred to as next-generation Layer 3 VPN multicast) constitute the next evolution after dual multicast VPNs (draft-rosen) and provide a simpler solution for administrators who want to configure multicast over Layer 3 VPNs.
The main characteristics of multiprotocol BGP-based multicast VPNs are:
They extend Layer 3 VPN service (RFC 2547) to support IP multicast for Layer 3 VPN service providers
They follow the same architecture as specified by RFC 2547 for unicast VPNs. Specifically, BGP is used as the control plane.
They eliminate the requirement for the virtual router (VR) model, which is specified in Internet draft draft-rosen-vpn-mcast, Multicast in MPLS/BGP VPNs, for multicast VPNs.
They rely on RFC-based unicast with extensions for intra-AS and inter-AS communication.
Multiprotocol BGP-based VPNs are defined by two sets of sites: a sender set and a receiver set. Hosts within a receiver site set can receive multicast traffic and hosts within a sender site set can send multicast traffic. A site set can be both receiver and sender, which means that hosts within such a site can both send and receive multicast traffic. Multiprotocol BGP-based VPNS can span organizations (so the sites can be intranets or extranets), can span service providers, and can overlap.
Site administrators configure multiprotocol BGP-based VPNs based on customer requirements and the existing BGP and MPLS VPN infrastructure.
Route Reflector Behavior in MVPNs
BGP-based multicast VPN (MVPN) customer multicast routes are aggregated by route reflectors. A route reflector (RR) might receive a customer multicast route with the same NLRI from more than one provider edge (PE) router, but the RR readvertises only one such NLRI. If the set of PE routers that advertise this NLRI changes, the RR does not update the route. This minimizes route churn. To achieve this, the RR sets the next hop to self. In addition, the RR sets the originator ID to itself. The RR avoids unnecessary best-path computation if it receives a subsequent customer multicast route for an NLRI that the RR is already advertising. This allows aggregation of source active and customer multicast routes with the same MVPN NLRI.