MPLS is based on the realization that small ATM cells are not needed in the core of modern networks, since modern optical networks (as of 2001) are so fast (at 10 Gbits/second and well beyond) that even full-length 1500 byte packets do not incur significant queuing delays.
At the same time, it attempts to preserve the traffic engineering and out-of-band control that made ATM attractive for deploying large scale networks.
MPLS works by encapsulating packets with a very simple MPLS header containing one (or more) 'labels' or 'tags', called a label stack, at the start of the packet. These MPLS labeled packets are forwarded along an MPLS tunnel, a virtual circuit, based on the contents of the labels.
First an MPLS tunnel is setup. To setup an MPLS tunnel, dynamic signaling protocols like RSVP-TE, LDP, CR-LDP are used, and in some case tunnels can be setup statically by configuration at each hop of the tunnel. Using these signaling methods an MPLS tunnel is setup between two routers. The entry point of an MPLS tunnel is called the ingress router, the end point is called the egress router and intermediate routers between the ingress and the egress are called transit routers. When a tunnel is setup, labels are exchanged between the egress, the transit and the ingress routers and with this the tunnel is ready to forward traffic.
At the ingress when a unlabelled packet is passed on to the MPLS tunnel, one (or more) label is inserted on the top of the packet and the packet is passed on the next hop router for this tunnel.
At a transit router, when a labeled packet is received, the topmost label is examined. Based on the contents of the label a swap, a push or a pop operation can be performed on the packets label stack. In a swap operation the label is swapped with a new label, in a pop operation the label is removed from the packet and on a push operation a new label is pushed on top of the existing label. After the label operations are performed the packet is forwarded to the next hop router for the tunnel.
During these operations, the contents of the packet below the MPLS Label Stack is not examined. The forwarding of the packet is done based on the contents of the labels, and so this allows protocol independent packet forwarding that does not need to look at a protocol-dependent routing table and avoid the expensive IP longest prefix match at each hop.
As the outermost label can be "popped" off, may reveal an inner label below, in a process called "decapsulation", or another label can be "pushed" on, "encapsulating" the packet in another layer of MPLS. This allows the hierarchical routing of MPLS packets.
At the egress when the last label has been popped, what remains is the payload, which can be an IP packet, or any of a number of other kinds of payload packet. In some special cases the last label can also be popped off at the penultimate hop (to hop before the egress), it is called PHP -Penultimate Hop Popping.
Labels are used to specify a forwarding equivalence class: that is to say, all packets with the same label are treated the same way for the purposes of packet forwarding.
However, MPLS can make use of existing ATM network infrastructure, as its labeled flows can be simply mapped to ATM virtual circuit identifiers, and vice-versa.
Unlike IP, MPLS does not define a directly usable end-point protocol. It only defines a way of encapsulating other layer 2 and layer 3 protocols. In this regard, it is similar to a protocol like PPP. Also unlike IP, MPLS explicitly decouples routing from forwarding, although it can fall back to using IP-style routing if necessary.
MPLS packets can be 1500 bytes long, plus any labels added to the packet. (Note that this requires the use of "baby jumbo packets" if Ethernet is used as the transport for MPLS). This compares well with the 48-byte cell of ATM, and reduces encapsulation overheads, particularly in the case of small packets: for example, it allows a minimum-length TCP packet to reside in a single MPLS packet, rather than two cells as in ATM.
The 16 bits of VCI and 8 bits of VPI in the ATM cell are replaced by a single label field of 20 bits, packed into a 32-bit label header. The 32-bit MPLS label field also contains an 8-bit time-to-live field, a "top of stack" bit, and three spare bits for expansion.
Although fewer bits are available for the label, labels can be stacked to create arbitrarily complex MPLS label stacks. This makes MPLS addressing and trunking vastly more flexible than that of ATM, as there is no need to impose an arbitrary boundary between VP and VC switching.
MPLS is currently in use in large networks, even though it is still (as of 2002) undergoing IETF standardization.
In practice, MPLS is mainly used to forward IP datagrams and Ethernet traffic. Major applications of MPLS are Telecommunications traffic engineering and MPLS-VPN.
L2TPv3 is emerging as a potential competitor to MPLS, particularly in existing networks with IP-only cores.How MPLS works
Comparison of MPLS versus IP
Comparison of MPLS versus ATM
MPLS deployment
Competitors to MPLS