Scalable Multicast Using MPLS in Software Defined Network

Abstract

Multicast helps to deliver data to multiple receivers efficiently. One scalability challenge faced by multicast is the per-channel forwarding states being maintained in the network layer, which increases linearly with the number of established multicast channels. MPLS helps to alleviate this problem by removing forwarding states from non-branch routers on the multicast tree and label switch packets in non-branch routers. To reduce the number of forwarding states in branch routers, many solutions were proposed to merge multicast trees/subtrees from different channels. Software Defined Network (SDN) decouples the control plane from the data plane, which enables low cost commodity design in routers and flexible network feature deployments through software implementation in centralized controllers. Equipped with SDN’s flexible policy and packet processing action installation, multicast tree/subtree merging becomes more convenient in SDN. This paper proposes a new scalable multicast solution in SDN to further reduce the number of forwarding states in routers. In the new solution, first a 2 level MPLS label switching scheme is used to reduce the extra point to point LSPs needed when multicast trees are merged. Secondly, a new multicast tree construction algorithm is designed to pursue more aggressive subtree matching between channels by taking advantage of per channel packet dropping actions in SDN. Simulation results show that the new solution can achieve 10-20 percent reduction in the number of forwarding entries needed for multicast traffic’s forwarding.