Cisco Multiservice Switching Networks

It is common practice to use RPM/B or RPM-PR cards in the MGX-8230 or MGX-8250 as PE routers. The Customer Edge (CE) routers connect to the PE using either Fast Ethernet (FE) interfaces in the RPM backcard or Frame Relay (FR), Point-to-Point Protocol (PPP), or ATM PVCs from FRSM or AUSM service modules.

The two designs to connect an MGX-8230 or MGX-8250 edge concentrator shelf to either a BPX-8600 or MGX-8850 routing switch, using AutoRoute and MPLS are as follows:

• Using separate physical interfaces for AutoRoute and MPLS (as we did in this chapter)

• Using only one physical link and a wraparound solution (a physical loop between two interfaces) in the BPX-8600 or MGX-8850 LSR

Using Separate Physical Interfaces

This design is straightforward. Basically, we use one physical port in the PXM-1 for the feeder trunk (running Annex.G) and separate physical ports for MPLS traffic. This design is shown in Figure 6-12.

This design is especially practical for colocated LSRs and eLSRs.

VPCs from the RPM cards to the PXM-1 port terminate in VNNI ports in the MGX-8850 or virtual trunks in the BPX-8600. The virtual path connections and VP-Tunnel LC-ATM interfaces cannot use the same interface because feeder trunks and VNNI/VTs cannot coexist in the same physical interface. This setup also provides the advantage of separating MPLS and other traffic. We can use only one feeder trunk but as many MPLS VP-Tunnel LC-ATM interfaces as needed.

Using One Uplink and a Wraparound

In some cases, the design just outlined is impractical, cost-prohibitive, or impossible. It might be impractical in situations where MGX-8200s (except MGX-8220, which doesn't support RPM cards) are in remote locations. This is impossible with OC-12 interfaces because the PXM-1 backcard has only one port. In these cases, you can use the design shown in Figure 6-13.

In this setup, the VPCs in the MGX-8230 or MGX-8250 from RPM cards terminate on the feeder trunk in the PXM-1. On the MGX-8850 AXSM port or BPX-8600 BXM port, we cannot have VNNIs or VTs in a feeder trunk. So we create VPCs to another port. These VPCs are PVPs (Permanent Virtual Paths) in a BPX-8600 under the AutoRoute control plane, or SPVPs (Soft Permanent Virtual Paths) in an MGX-8850 under the PNNI control plane.

This port where the VPCs terminate is physically looped back to the port where the VNNI or VT is configured.

This design has one further benefit. As shown in Figure 6-14, only one physical loop is needed per LSR for multiple MGX-8230 and MGX-8250 edge concentrator shelves, as long as bandwidth and LCN permit.

Even though only two MGX-8230 or MGX-8250 feeder shelves are shown, multiple MGX edge concentrators are supported with the same model.