Layer 2 Interworking

In the previous section, we discussed at length the like-to-like model with the same transport type at each end. This like-to-like AToM model can easily be extended with the interworking of disparate transport typesthat is, different transport type at each end. For example, one attachment circuit could be Ethernet and another could be Frame Relay VC. This type of interworking is not unknown to the industry. RFC 2684 and RFC 2427 are standard documents that specify how to encapsulate and deliver Ethernet and IP frames over ATM and Frame Relay VCs, respectively. With these models, customer devices can be interconnected using a Frame Relay VC or ATM VC, and Ethernet frames or IP frames can be transported across these devices. The interworking function is performed at the device where the VC is terminated, and the frames are either bridged nor routed to the destination.

Let us consider a case study to determine why this type of Layer 2 interworking can be an important service. Consider an enterprise with as many as 1000 branch sites that need to be connected to a head office site and a disaster recovery site. A traditional hub-and-spoke design for a Layer 2 VPN requires that at least one Layer 2 connection be terminated from each of the branch sites to the head office site and sometimes another one to the backup site. Because the enterprise is managing the IP routing, it might want to buy redundancy in the wide area network (WAN) and sign up for a strict SLA. The offer it gets from the service provider for this Layer 2 VPN might not be attractive if the enterprise is asked to terminate 1000 DLCIs or ATM VCs at the head office and backup site from the remote sites. Managing a large number of DLCIs or ATM VCs at the head office and branch sites might be an issue. Moreover, it might be costly because the charge of the circuit is usually per virtual connection.

Instead, the enterprise customer might ask for Ethernet connectivity at the head office and Frame Relay or ATM VCs at the remote sites. This is a Layer 2 service, so it implies that the Frame Relay or ATM frames from the branch sites needs to be translated to Ethernet frames at the head office. This requires the service provider to translate protocols from one side of the network to another.

The challenge of translating protocols from one to another is not trivial. The characteristics of the protocols, such as point-to-point versus broadcast, make it a challenging task. Routing protocols behave differently over broadcast medium such as an Ethernet as opposed to over a point-to-point link, such as FR DLCI or an HDLC link. Hence, translating between broadcast protocols, such as Ethernet, and point-to-point protocols, such as Frame Relay or ATM, require additional configurations, not just on the PE devices, but also on the CE devices. For example, route bridge encapsulation (RBE) configuration can be required if bridged interworking is configured.

Interworking Modes

The two main Interworking modes are bridged interworking and routed interworking. Sometimes these two modes are also referred to as Ethernet interworking and IP interworking.

Ethernet interworking or bridged interworking allows delivery of Ethernet frames across the two transport types. This requires the capability to use the pseudowire as a bridge for Ethernet frames between the different transport types at each end. The CPE configuration required in this case is no different from like-to-like service.

Routed interworking or IP interworking allows the delivery of IP frames across the various transport types. This is a common service where IP routers or hosts are connected across the interworking pseudowire. For example, a router is connected to a point-to-point HDLC link on one side of the network and another router is connected with an Ethernet connection to the same network. Now these two routers need to establish a Layer 2 adjacency so that routing information can be exchanged. Using IP interworking, these two routers can peer with each other across the packet network despite the fact that each uses a different transport type for physical connectivity. Address resolution is done from IP to MAC using a proxy ARP function.

Interworking Models and Applications

With five transport types, several combinations of Layer 2 interworking are possible, including the following:

• Ethernet-to-Frame Relay interworking

• Ethernet-to-ATM interworking

• ATM-to-Frame Relay interworking

• Ethernet-to-PPP/HDLC interworking

• Frame Relay-to-PPP/HDLC interworking

• ATM-to-PPP/HDLC interworking

All these models have two modes, such as Ethernet interworking or IP interworking. Depending on the service model, any of these modes can be used. These combinations provide flexibility in the connectivity of different sites in a Layer 2 VPN. For example, the problem described at the beginning of this section can easily be addressed by interworking remote DLCIs with Ethernet VLANs, avoiding termination of 1000 DLCIs at the head office or disaster recovery site.

Assuming an operational MPLS network (LDP or MPLS TE is enabled appropriately in the network), the configuration requirements in this type of Layer 2 Interworking are as follows:

1. Configure the CPEs (both sides) with the attachment circuit information, including VLAN IDs, DLCI values, and ATM VPI/VCI or PPP session information.

2. Configure PEs (PE1 and PE2) with the same AC information (AC1 on PE1 and AC2 on PE2).

3. Configure the PE1 with the cross-connect information with the AC1 and pseudowire and specify whether the PE should do the interworking.

4. Configure QoS parameters on the AC and the pseudowire.

5. Map the pseudowire to the TE tunnel if bandwidth guarantees are necessary or you're using MPLS TE in the network.

6. Configure PE2 with the cross-connect information with AC2 and the pseudowire. Similar to PEI, you must specify whether interworking is required.

7. Configure QoS information on PE2 for the AC2 and pseudowire.

8. Configure the CPE with RBE/IRB capability depending on the interworking mode.

The operational complexity of this configuration is higher than that of an existing ATM/FR Layer 2 network, but the flexibility this mode provides can offset some operations costs associated with it. Refer to the section titled "Benefits of L2VPNs" for more details.