A scaling problem inherent in a full-mesh VPLS network is packet replication.
In a full-mesh configuration, until a node learns over which pseudowire a MAC address is reachable, unknown unicast frames must be flooded on all pseudowires within the VPLS. Packet replication is always true for broadcast and multicast traffic. As the number of VPLS peers increase, the packet replication burden on a node increases. MTU devices attached to a full-mesh core most likely cannot maintain wire-speed forwarding as the number of VPLS peers increase. Hierarchical VPLS eliminates this MTU burden by requiring only a single pseudowire connection between a spoke and its core PE peer. Packet replication is pushed to the PEs, where it is more suitably handled.