Control plane and data plane

The CNX exchange fabric is the data plane. A packet sent from one member to another crosses the shared Layer-2 fabric directly between their routers. CNX route servers are not in that forwarding path.

The route servers operate in the control plane. They receive BGP routes from members, validate them, apply supported community controls, and redistribute accepted paths. Their ASN, AS132213, is not inserted into the redistributed AS_PATH.

                         BGP control plane
Member A router  ─────── CNX route servers ───────  Member B router
       │                                                  │
       └────────── packets on the Layer-2 fabric ─────────┘

Each member's router applies its own import policy and BGP decision process. The resulting forwarding table determines which next hop carries traffic. CNX does not set a member's local preference. CNX recommends giving accepted CNX routes a higher local preference than paid transit; each member retains control of how CNX and bilateral routes compare.

Route-server and bilateral sessions

Route-server peering provides reachability to participating members without a separate BGP session for every relationship. Members remain free to establish bilateral sessions over the peering LAN.

When a prefix is learned through both models, local policy decides which path wins. A common design gives a bilateral session a different local preference from the route-server session, but that choice belongs to the member.

Loss of a route-server session is a control-plane event. Existing forwarding continues until routes are withdrawn or expire. It does not mean that packets were traversing the failed route server.