Joint Multicast Application-Group Mapping and Rendezvous Point Selection for Multi-Tenancy Data Center Networks

Modern cloud applications frequently reside in overlay networks and exhibit one-to-many communication patterns. The Ingress Replication (IR) approach can support overlay multicast traffic by replicating packets at the ingress node and sending a unicast copy of the packet to each interested receiver. However, IR consumes more bandwidth as multiple copies of the same packet are forwarded through the network. The tenant-routed multicast architecture addresses this drawback by leveraging IP multicast in the underlay network to transport overlay multicast application traffic. The overlay multicast traffic is encapsulated with an underlay multicast group address and then transported across the network without duplication. However, the limited number of underlay multicast groups makes it challenging to efficiently support a massive number of multicast applications, which is typical in cloud environments. In this paper, we address the problem of mapping overlay applications to underlay multicast groups and selecting Rendezvous Points (RPs) to minimize network utilization. We provide a mathematical formulation for this joint problem and design a local search-based algorithm to solve it. Extensive experimental results demonstrate that our solution can approximate the exact solution and significantly reduce network utilization compared to conventional thresholdbased methods.