Multi Protocol Label Switching Introduction

Multi Protocol Label Switching (MPLS) is a mechanism that when added to a routed IP network allows packets to be switched in hardware at high speed. It is considered a Layer 2 technology, although in many circles it is referred to as Layer 2 and a half as it mainly functions between the Data-Link Layer and the Network Layer of the OSI Reference Model.

A number of vendors had similar but proprietary technologies that became the basis for MPLS, for instance Cisco had Tag Switching and IBM had similar technology known as ARIS (Aggregate Route Based IP Switching). In 1997 a working group was formed under the guidance of the IETF (Internet Engineering Task Force), and eventually the name MPLS was decided upon.

An MPLS label is a short 32 bit identifier which is used to switch packets within an MPLS domain. When used with ATM (Asynchronous Transfer Mode) the label replaces the VPI/VCI, and a similar situation occurs when MPLS is used with some other Data-Link Layer technologies such as X.25 and Frame Relay. When Multi Protocol Label Switching is used with Ethernet, PPP, FDDI or Token Ring, the 32-bit Label is inserted between the Data-Link Layer and Network Layer headers. A label used in this manner is often referred to as a 'Shim Label'. Labels are normally changed switch by switch as the packet traverses the LSD (Label Switched Domain), and this is where the term Label Swapping comes from.

The original concept of MPLS was to combine the intelligence of Layer 3 Routing with the speed of Layer 2 Switching in order to provide a high throughput of Layer 3 packets. The technology is maturing and packet flows with the same FEC (Forward Equivalence Class) can be switched across certain paths to afford that packet flow the appropriate Quality of Service.

Let us take a look at some of the terminology:

An LSP (Label Switched Path) is a path set up across a network from the entry point known as the ingress to the exit point known as the egress. The path is defined in forwarding tables held in the switching devices, with the patch being a series of hops across the label switched domain. Any data which conforms to an exact match algorithm performed upon the IP header information, referred to as a Forward Equivalence Class will be forwarded across the LSP.

An LER (Label Edge Router) is a device at the ingress or egress which applies the first label to the packet and removes the last. An edge LSR receiving an unlabelled packet will determine the appropriate FEC using conventional forwarding information within the IP packet. Label switching and forwarding will be used for the hop to the next Label Switching Router. If an edge LSR receives a labelled packet with a non LSR destination the label is stripped and the packet forwarded using only the conventional routing components.

An LSR (Label Switching Router) performs the routing and switching within the LSD. Certain Layer 2 switches can apply Layer 3 intelligence and Routers can have Layer 2 switches associated with them, the net result being devices that can switch labelled packets and route unlabelled packets.

Any LSR node must have a forwarding table populated with the correct information. This table is known as the LIB (Label Information Base), and its role is to provide the information required to bind each FEC to the appropriate LSP. Some mechanism must be used to populate each LIB, and this is normally performed by a LDP (Label Distribution Protocol). An alternative to this is to piggyback the information onto a standard routing protocol such as BGP (Border Gateway Protocol).

The MPLS Labels themselves are 32 bits in length and are comprised:

A 20-bit value

3 Experimental bits now often used as the Traffic Class field for Quality of Service

An 1-bit flag that is used to indicate the bottom of the stack when multiple labels are loaded onto the same packet flow.

An 8-bit TTL (Time-to-Live) field, used in a similar manner to the TTL in an IP packet.

In the case of data link layers which add a shim label between Layer 2 and Layer 3, labels can be stacked one behind the other so that LSPs only have to action the top label and then remove it or 'Pop it' before forwarding the packet. In ATM the label represents the VCI/VPI. The TTL will be decremented to indicate the number of LSRs through which it has been forwarded.

Because routes can be deterministic, MPLS has evolved and can be used to create a number of different VPN types, the most common of which is the Point-to-Point VPN which is often referred to as a Pseudowire or Virtual Leased Line.

This article only provides basic information on the subject, but at least it provides the reader with the concept of MPLS.

This article on MPLS was written by David Christie, MD at NSTUK Ltd, Website http://www.nstuk.com