MPLS

t the end of 2001, MPLS (MultiProtocol Label Switching) was the subject of a large number of articles and conferences, as well as a growing number of announcements from network equipment manufacturers. Now that the first commercial services based on an MPLS/IP core network are appearing, the benefits of the technology seem to have been demonstrated by their smooth operation. However, a thorough understanding of MPLS is still essential to be able to make the right choices. That's why we need to look beyond the hype and understand the reasons behind the definition of MPLS, and the real contribution of MPLS and related technologies to modern network cores.

The primary principle of label switching is to replace the time-consuming and complex processing associated with IP packet relaying with simpler processing. From another point of view, the aim is to use IP signaling to control circuit-switched networks, enabling the integration of different Layer 2 switching technologies in the same core network. In a conventional IP network, each router decides, on the basis of the destination address contained in a packet's header, whether it is destined for one of the directly connected subnetworks or, if not, to which neighboring router it should forward the packet. To make this decision, it uses the contents of its routing table, which is constructed by the routing protocols. This table associates network and subnet addresses, or more generally IP address prefixes, with the next router on the path to the destination network. As this prefix can be of variable length, and as the order is not imposed in the routing table, the router must examine the entire routing table to decide which table entry best matches the packet's destination address. This process is relatively costly, given the ever-increasing size of the routing tables at the heart of the Internet (around 100,000 entries in 2001).

Label switching greatly reduces the cost of this search, as it is no longer carried out by the core network equipment. In fact, each piece of equipment within the core network performs the search only once, when the path is created. To this end, a label is added to each packet by the routers at the domain border. They can choose the label according to various criteria, including the packet's destination address. Relay decisions by equipment in the network core are made on the basis of the label added to the packet. All packets with the same label are treated in the same way by the label switches. They undergo the same relaying treatment, and a fortiori, follow the same path to the label-switched network output. The equipment has a switching table containing, for each incoming label, the output interface and an outgoing label.

The same label can be associated with packets...