PBB-EVPN (Informational)
Last updated
Last updated
Back when EVPN was new, Cisco first appeared to only support PBB-EVPN. I suppose it was thought that PBB has much better scale, because only the core MACs are advertised in BGP. If there are millions of VMs to handle in the customer bridge domain, this might be too much for BGP.
The idea is fairly simple. The core consists of a bridged L2 network. The edge devices (PEs), instead of using an MPLS label for transport, use an additional L2 header with the destination MAC as the egress PE. This is MAC-in-MAC, also known as 802.1ah.
The benefit of MAC-in-MAC is primarily scalability. The core only has to switch on core MACs. The customer traffic is tunneled inside the core L2 switched network.
For this to work, the ingress edge (BEB or backbone edge bridge in 802.1ah terms) encapsulates the customer packet with an outer header that contains the egress BEB’s MAC and an instance ID. This instance ID is essentially a VPN ID. The egress BEB decapsulates the outer header, and knows which customer bridge domain the frame belongs to due to the instance ID. This is called the I-SID (Instance service ID).
When you add EVPN to this, EVPN is only used for the provider L2 network. So only backbone MAC addresses are advertised. You are essentially just using PBB but overlaying it on a modern routed/MPLS core network.
The configuration of this is fairly basic. You define a separate core bridge-domain and a customer bridge-domain. The customer BD points to the core BD which enables 802.1ah encapsulation. The EVI is only associated with the backbone BD. You define the PE’s backbone MAC manually as well.
This does not appear to be on the CCIE-SP topics list anymore. Bridge domains are not supported on the XRv9K anyways. So I don’t think we really need to go much further than this.
This blog post goes into more details and has lots of reference links for further reading: https://edkimura.blogspot.com/2017/11/mpls-l2vpn-pbb-evpn.html
