Definitive MPLS Network Designs [Electronic resources]

Layer 2 Services and Pseudowires

Traditionally, the components of data networks consisted of a number of layers, each providing a well-defined service to the layers above them. At the Transport layer, service providers could deploy optical equipment (such as SDH/Sonet/xWDM) and copper (xDSL, E1, T1) technologies to present leased-line services. Above this layer, Transmission services were deployed (X.25, Frame Relay, ATM, and so on). These provided virtual leased line or virtual circuit (VC) services. The last layer, the Network layer, provided IP services such as Internet access and network-based Layer 3 VPNs. Today, most routers can bypass the transmission layer and obtain services directly from the transport equipment.

Due to the high demand for Layer 3 services, the capabilities in terms of capacity on routers have outstripped those at Layer 2, where most providers have deployed OC-12 links in their ATM backbones but OC-48 or even OC-192 in their IP/MPLS network. Because of this, these same providers are considering using the bandwidth available on their IP networks to deliver Layer 2 services. This is indeed what has happened in the recent past with the introduction of pseudowire technology based on MPLS.

Pseudowire technology provides all the mechanisms to emulate a Layer 2 transport service over an IP/MPLS infrastructure.

Pseudowire Network Components

[pwe3-req]. These different elements provide different functions:

pwe3-cp] relies on an LSP (called the tunnel LSP) in each direction between two pseudowire-capable edge devices. An MPLS label identifies the tunnel LSP and is used to carry one or multiple pseudowires between a given pair of ingress and egress PE devices. Tunnel LSPs may be created by distributing labels automatically using LDP or BGP with label distribution, by deploying traffic-engineered tunnels between PE devices, or by setting static labels along the path between the source and destination PE devices.

Directed LDP session This targeted LDP session runs between PE devices that provide pseudowire services. The session is used to signal pseudowire setup and status.

Virtual circuit label The pseudowire Forwarding Equivalence Class (FEC) provides a unique locally assigned label for each AC. These labels are signaled across the directed LDP session to the appropriate remote PE device.

Pseudowire identifier (PWid) Each pseudowire is identified through a unique identifier (32 bits).

Some of these components are shown in Figure 1-27.

Note

Figure 1-27 shows that multiple pseudowires may be carried across the tunnel LSP. A single tunnel LSP is adequate because the virtual circuit label identifies the individual pseudowires.

Pseudowire Forwarding Equivalent Class

Pseudowire setup is signaled between PE devices using LDP downstream unsolicited mode across a directed session. As illustrated in Figure 1-28, an LDP label mapping message is used to convey the information for a given pseudowire. It contains a FEC TLV, a label TLV, and optional parameter TLVs (if necessary).

As shown in Figure 1-29, the FEC TLV contains a PWid FEC element that identifies the pseudowire circuit associated with the advertised label.

Pseudowire Creation and Signaling

There are several ways that two attachment circuits may be cross-connected to establish a pseudowire circuit between them. [pwe3-cp] describes the use of automatic discovery and automatic configuration. However, the most common method at this time is manual configuration through the router's command-line interface (CLI).

Using this method, the operator must initially establish the attachment circuits at each of the PE routers through which the pseudowire will connect. Having established the local attachment circuits, a pseudowire connection may be configured, including the endpoints of the directed LDP session, unique PWid, and so forth.

If a directed LDP session between the ingress and egress PE routers already exists, signaling of the pseudowire may take place using this existing session. However, if one does not exist, it is created based on the endpoint parameters (such as IP addresses) specified in the pseudowire setup configuration.

After receiving the LDP label mapping message, each PE router can match the PWid information and therefore establish the connection. The label carried within the message for this pseudowire is set on all packets flowing for this circuit between ingress and egress PE routers.

This sequence of events is shown in Figure 1-30.

Pseudowire Encapsulation

[pwe3-atm], [pwe3-fr], [pwe3-eth], and [pwe3-sonet]. In addition to this encapsulation, a control word (32-bit entity) may be carried. It provides additional information for a given service. For example, the control word is used with Ethernet encapsulation to ensure sequencing. With Frame Relay, the control word contains protocol control information such as Forward Explicit Congestion Notification (FECN), Backward Explicit Congestion Notification (BECN), Discard Eligibility (DE), and so forth, as well as sequencing.

Pseudowire Packet Flow

After the pseudowire is successfully established, each Layer 2 Protocol Data Unit (PDU) that enters the PE router on an incoming attachment circuit associated with the pseudowire is sent to the pseudowire.

Figure 1-31 shows how the Layer 2 packet is carried across the MPLS network with an inner label (called the VC label and shown as VL in the figure) that identifies the pseudowire, and an outer label (shown as OL in the figure) that identifies the tunnel LSP between the ingress and egress PE routers.