The point at which an interconnection between two different MPLS networks is made is referred to as an MPLS Exchange Point (MEP). A straightforward MEP would be at the customer premises.
The customer must know what the COS markings mean on each side of the MEP and how to configure the interconnection router to apply them. The customer (or their consultant) also must know how to use BGP to interconnect his router with the provider-edge routers in both networks and propagate the necessary VPN routing information.
The MEP router de-encapsulates the VPN data from its MPLS headers in network A and applies the appropriate COS and MPLS encapsulation for egress into network B. While packets belonging to each COS are between networks A and B they are kept separate at the link layer within the MEP router.
If multiple MPLS VPNs are being interconnected, their data must be kept segregated from each other. There are a number of techniques for isolating this traffic, including:
- Using link layer traffic isolation mechanisms, such as virtual local area networks (VLANs), frame relay, or circuit cross-connect.
- Using a separate MEP router for each MPLS-VPN instance.
Interconnecting MPLS networks requires subject matter expertise. Network professionals must understand MPLS internals, vendor-specific switch/router implementations and carrier-specific COS markings. Additionally, the enterprise must be willing to build MEPs at points where carrier MPLS networks come together—typically carrier-neutral collocation facilities, or pay for local loops to the enterprise facility. Unless you have the expertise in-house, it makes sense to outsource this configuration and management.