Traditional IP routing has several well-known limitations, ranging from scalability issues to poor support of traffic engineering and poor integration with Layer 2 backbones already existing in large service provider networks. With the rapid growth of the Internet and the establishment of IP as the Layer 3 protocol of choice in most environments, the drawbacks of traditional IP routing became more and more obvious.

MPLS was created to combine the benefits of connectionless Layer 3 routing and forwarding with connection-oriented Layer 2 forwarding. MPLS clearly separates the control plane, where Layer 3 routing protocols establish the paths used for packet forwarding, and the data plane, where Layer 2 label switched paths forward data packets across the MPLS infrastructure. MPLS also simplifies per-hop data forwarding, where it replaces the Layer 3 lookup function performed in traditional routers with simpler label swapping. The simplicity of data plane packet forwarding and its similarity to existing Layer 2 technologies enable traditional WAN equipment (ATM or Frame Relay switches) to be redeployed as MPLS nodes (supporting IP routing in the control plane) just with software upgrades to their control plane.

The control component in the MPLS node uses its internal data structure to identify potential traffic classes (also called Forward Equivalence Classes). A protocol is used between control components in MPLS nodes to exchange the contents of the FEC database and the FEC-to-label mapping. The FEC table and FEC-to-label mapping is used in Edge-LSRs to label ingress packets and send them into the MPLS network. The Label Forwarding Information Base (LFIB) is built within each MPLS node based on the contents of the FEC tables and the FEC-to-label mapping exchanged between the nodes. The LFIB then is used to propagate labeled packets across the MPLS network, similar to the function performed by an ATM switching matrix in the ATM switches.

The MPLS architecture is generic enough to support other applications besides IP routing. The simplest additions to the architecture are the IP multicast routing and quality of service extensions. The MPLS connection-oriented forwarding mechanism together with Layer 2 label-based look ups in the network core also has enabled a range of novel applications, from Traffic Engineering to real peer-to-peer Virtual Private Networks.

    Part 2: MPLS-based Virtual Private Networks