The VXLAN spec allows for two implementation types. One uses multicast. What is the other one called?
The VXLAN spec allows for two implementation types. One uses multicast. What is the other one called?
The VXLAN specification allows for two implementation types: multicast and head-end replication. Head-end replication, also known as ingress replication, is an approach where the source VXLAN Tunnel Endpoint (VTEP) replicates traffic and sends it directly to the destination VTEPs without needing multicast support in the underlying network.
there is only head-end replication. Head-end replication is a data distribution method that supports broadcast, unknown unicast, and multicast (BUM) traffic over VXLANs by replicating BUM data locally for transmission to the set of remote VTEPs specified by a flood list. This data flooding facilitates remote MAC address learning through the forwarding of data with unknown MAC addresses.
https://eos.arista.com/tag/head-end-replication/
Unicast: As of EOS-4.13.0F, Arista supports unicast-based VXLAN (no multicast required) to support broadcast, unkown unicast and multicast traffic. This is called Head-End replication (HER)
There is no BUM replication, but BUM traffic. HER is triggered every time there's BUM traffic.