What Is Multiprotocol Label Switching?

Multiprotocol label switching (MPLS) is a packet forwarding technique in which routers direct data along predetermined paths using short fixed-length labels rather than performing a full destination-address lookup at each hop. A label is prepended to each packet at the network ingress, routers along the path swap or pop that label to determine the outgoing interface, and the label is removed at the egress. The "multiprotocol" qualifier reflects that MPLS is applicable to any network layer protocol: it was originally designed for IP but can encapsulate ATM cells, Ethernet frames, and other payload types. The architecture is defined in RFC 3031, published by the IETF, which introduced the label-switching paradigm and the terminology of label edge routers (LERs) and label switch routers (LSRs).

MPLS emerged in the late 1990s as a way to combine the traffic engineering capabilities of Asynchronous Transfer Mode (ATM) virtual circuits with the scalability of IP routing. ATM had demonstrated that connection-oriented forwarding using short labels could achieve deterministic performance guarantees, but ATM's cell-based design was cumbersome for IP traffic. MPLS adopted the label concept and applied it at the IP layer, producing a technology that could be overlaid on existing IP infrastructure.

Label Distribution and Routing Protocols

Before labels can be applied, routers must establish a mapping between network destinations and label-switched paths (LSPs). The Label Distribution Protocol (LDP) automates this by propagating bindings between IP prefixes and labels through the network. Resource Reservation Protocol with Traffic Engineering extensions (RSVP-TE) allows explicit routing of LSPs with bandwidth and priority constraints, enabling traffic engineering that distributes load away from congested links. These signaling protocols interact with IP routing protocols such as OSPF and IS-IS, which provide the topology information that path computation uses. The integration of MPLS with routing protocols is covered extensively in IEEE publications on MPLS traffic engineering.

Traffic Engineering and VPN Services

Traffic engineering is one of the two principal applications of MPLS in deployed networks. By establishing explicit LSPs with reserved bandwidth, network operators can route flows along underutilized paths that a shortest-path routing protocol would not select. This capability is critical in large ISP backbone networks, where link utilization imbalances are common. The second major application is virtual private network (VPN) services. MPLS VPNs, defined in RFC 4364, allow service providers to offer customers isolated layer-3 connectivity across a shared infrastructure by using a second label in a two-level label stack: an outer transport label identifies the path to the egress PE router, and an inner VPN label identifies the customer's forwarding table. Together, these capabilities make MPLS the primary transport technology in carrier backbone networks, as described in IETF RFC 3031 on MPLS architecture.

Segment Routing and Evolution

Segment routing (SR) is a more recent forwarding paradigm that extends MPLS label stacks to encode explicit paths through a network without per-flow state in intermediate nodes. In SR-MPLS, path instructions are encoded as a sequence of segment identifiers (SIDs), each mapped to a label, pushed at the ingress node. This simplifies the control plane by eliminating RSVP-TE state from the core, while retaining the ability to steer traffic along explicit routes. SR-over-IPv6 (SRv6) replaces the MPLS label stack with IPv6 segment routing headers, extending the paradigm to pure IP environments.

Applications

Multiprotocol label switching has applications in a wide range of fields, including:

• Carrier backbone traffic engineering and load balancing
• Enterprise and carrier MPLS VPN services for customer isolation
• Quality-of-service differentiation for voice, video, and data traffic
• Pseudowire emulation of TDM, ATM, and Ethernet circuits over IP networks
• Data center wide-area network interconnect with deterministic path control