On forwarding state control in VPN multicast based on MPLS multipoint LSPs

Abstract

The demand for multicast-capable VPN services, like Virtual Private LAN Service (VPLS), has grown quickly in the last years. In order to save bandwidth, MPLS point-to-multipoint LSPs could be used, but the VPN-specific state information to be handled inside the network may exceed the capacity of core nodes. A well-known solution for this is to aggregate the multicast/broadcast traffic of multiple VPNs into shared p2mp LSP trees. In shared trees, although some bandwidth is wasted because a fraction of the packets are delivered to non-member leaves (either not in the VPN broadcast or multicast group), there is wide working range where a good state vs. bandwidth trade-off is achieved. In this paper we enhance and improve previous works that analyze this trade-off. We propose new techniques for multicast traffic aggregation of VPNs in MPLS-based networks, with the objective of observing the behavior of the aggregation philosophy for different aggregation degrees, which should be very useful for network design and deployment purposes. We assess the aggregation heuristics over different reference networks and VPN geographic distributions. Simulations give a quantitative indication of the relevance of intelligent aggregation, of geographical distribution and group sizes.