Layer2 Bridge support in Multi POP Deployments

Layer2 Bridge support in Multi POP Deployments

This feature is available from Release 1.0.1 onwards.

This feature is applicable to Layer 2 bridging and when Deterministic Prefix Allocation (DPA) is configured on the e2e controller.

In terragraph network, CNs and DNs are allocated prefixes from a seed prefix. There are various way of prefix allocation. In DPA, controller assigns prefixes to nodes based on the network topology to allow POP nodes to take advantage of route summarization and help load balance ingress traffic.

CNs & DNs see multiple POP node in mesh, they select POP to form GRE tunnel, by matching there lo ipv6 address with POPs lo ipv6 address. The longest prefix match is selected as best POP for L2 GRE Tunnel establishment.

Multi-POP setup gives the advantage that user data traffic can take alternate routes if the best route is unavailable for some reason. Open/R make this selection to route the traffic.
If due to some reason best pop is unavailable, CNs & DNs switch to next best POP. They however keep track of their primary POP availability and switch to it once it become online

Is there any chance that using a using a multi pop deployment with layer 2 bridging enabled could lead to a broadcast storm and related spanning tree issues? The reason I ask is that the POP nodes have layer 2 connectivity between them, plus all the end user traffic is landed into the same layer 2 broadcast domain. So what prevents this broadcast traffic from not being forwarded back through the network via a different POP?

Assuming there is no wired connection between CNs / DNs (other than ones marked as relay port on both ends), broadcast traffic will not be forwarded back and cause a loop.
I will explain the case of Downlink broadcast packets
Tunnels are one to one between POP to CNs/DNs. Over tunnel these packets reach CNs/DNs and they are bridged out only on ethernet ports (that are not marked as relay port). So the packets don’t have a way back.

1 Like