The Multiprotocol Label Switching labels are advertised between routers so that they can build a label-to-label mapping. These labels are attached to the IP packets, enabling the routers to forward the traffic by looking at the label and not the destination IP address. The packets are forwarded by label switching instead of by IP switching. The label switching technique is not new. Frame Relay and Asynchronous Transfer Mode use it to move frames or cells throughout a network. In Frame Relay, the frame can be any length, whereas in Asynchronous Transfer Mode, a fixedlength cell consists of a header of 5 bytes and a payload of 48 bytes. The header of the Asynchronous Transfer Mode cell and the Frame Relay frame refer to the virtual circuit that the cell or frame resides on.

The similarity between Frame Relay and Asynchronous Transfer Mode is that at each hop throughout the network, the “label” value in the header is changed. This is different from the forwarding of IP packets. When a router forwards an IP packet, it does not change a value that pertains to the destination of the packet; that is, it does not change the destination IP address of the packet. The fact that the Multiprotocol Label Switching labels are used to forward the packets and no longer the destination IP address have led to the popularity of Multiprotocol Label Switching. These benefits—such as the better integration of IP over Asynchronous Transfer Mode and the popular Multiprotocol Label Switching virtual private network (Virtual Private Network) application.

Benefits of Multiprotocol Label Switching

This section explains briefly the benefits of running Multiprotocol Label Switching in your network. These benefits include the following:

■ The use of one unified network infrastructure
■ Better IP over Asynchronous Transfer Mode integration
■ Border Gateway Protocol (Border Gateway Protocol)-free core
■ The peer-to-peer model for Multiprotocol Label Switching Virtual Private Network
■ Optimal traffic flow
■ Traffic engineering Consider first a bogus reason to run Multiprotocol Label Switching. This is a reason that might look reasonable initially, but it is not a good reason to deploy Multiprotocol Label Switching.

Bogus Benefit

One of the early reasons for a label-swapping protocol was the need for speed. Switching IP packets on a CPU was considered to be slower than switching labeled packets by looking up just the label on top of a packet. A router forwards an IP packet by looking up the destination IP address in the IP header and finding the best match in the routing table. This lookup depends on the implementation of the specific vendor of that router. However, because IP addresses can be unicast or multicast and have four octets, the lookup can be complex. A complex lookup means that a forwarding decision for an IP packet can take some time. Although some people thought that looking up a simple label value in a table rather than looking up the IP address would be a faster way of switching packets, the progress made in switching IP packets in hardware made this argument a moot one. These days, the links on routers can have a bandwidth up to 40 Gbps. A router that has several high-speed links would not be able to switch all the IP packets just by using the CPU to make the forwarding decision.

The CPU exists mainly to handle the control plane. The control plane is the set of protocols that helps to set up the data or forwarding plane. The main components of the control plane are the routing protocols, the routing table, and other control or signaling protocols used to provision the data plane. The data plane is the packet forwarding path through a router or switch. The switching of the packets—or the forwarding plane—these days is done on specifically built hardware, or application-specific integrated circuits (ASIC). The use of ASICs in the forwarding plane of a router has led to IP packets being switched as fast as labeled packets. Therefore, if your sole reason for implementing Multiprotocol Label Switching in your network is to pursue the faster switching of packets through the network, it is a bogus reason.

The Use of One Unified Network Infrastructure

With Multiprotocol Label Switching, the idea is to label ingress packets based on their destination address or other preconfigured criteria and switch all the traffic over a common infrastructure. This is the great advantage of Multiprotocol Label Switching. One of the reasons that IP became the only protocol to dominate the networking world is because many technologies can be transported over it. Not only is data transported over IP, but also telephony. By using Multiprotocol Label Switching with IP, you can extend the possibilities of what you can transport. Adding labels to the packet enables you to carry other protocols than just IP over an Multiprotocol Label Switching-enabled Layer 3 IP backbone, similarly to what was previously possible only with Frame Relay or Asynchronous Transfer Mode Layer 2 networks. Multiprotocol Label Switching can transport IPv4, IPv6, Ethernet, High-Level Data Link Control (HDLC), PPP, and other Layer 2 technologies.

The feature whereby any Layer 2 frame is carried across the Multiprotocol Label Switching backbone is called Any Transport over Multiprotocol Label Switching (AToM). The routers that are switching the AToM traffic do not need to be aware of the Multiprotocol Label Switching payload; they just need to be able to switch the labeled traffic by looking at the label on top of it. In essence, Multiprotocol Label Switching label switching is a simple method of switching multiple protocols in one network. You need to have a forwarding table consisting of incoming labels to be swapped by outgoing labels and a next hop. In short, AToM enables the service provider to provide the same Layer 2 service toward the customers as with any specific non-Multiprotocol Label Switching network. At the same time, the service provider needs only one unified network infrastructure to carry all kinds of customer traffic.

Better IP over Asynchronous Transfer Mode Integration

In the previous decade, IP won the battle over all other networking Layer 3 protocols, such as AppleTalk, Internetwork Packet Exchange (IPX), and DECnet. IP is relatively simple and omnipresent. A much-hyped Layer 2 protocol at the time was Asynchronous Transfer Mode. Although Asynchronous Transfer Mode as an end-toend protocol—or desktop-to-desktop protocol—as some predicted, never happened, Asynchronous Transfer Mode did have plenty of success, but the success was limited to its use as a WAN protocol in the core of service provider networks. Many of these service providers also deployed IP backbones. The integration of IP over Asynchronous Transfer Mode was not trivial. To better integrate IP over Asynchronous Transfer Mode, the networking community came up with a few solutions.

One solution was to implement IP over Asynchronous Transfer Mode according to the well-known RFC 1483, “Multiprotocol Encapsulation over Asynchronous Transfer Mode Adaptation Layer 5,” which specifies how to encapsulate multiple routed and bridged protocols over Asynchronous Transfer Mode adaptation Layer (AAL) 5. In this solution, all Asynchronous Transfer Mode circuits had to be manually established, and all mappings between IP next hops and Asynchronous Transfer Mode endpoints had to be manually configured on every Asynchronous Transfer Mode-attached router in the network. Another method was to implement LAN Emulation (LANE). Ethernet had become a popular Layer 2 technology at the edge of the network, but it never achieved the scalability or reliability requirements of large service provider networks. LANE basically makes your network look like an emulated Ethernet network. This means that several Ethernet segments were bridged together as if the Asynchronous Transfer Mode WAN network in the middle were an Ethernet switch. Finally, Multiprotocol over Asynchronous Transfer Mode (MPOA), which is a specification by the Asynchronous Transfer Mode Forum, gives you the tightest integration of IP over Asynchronous Transfer Mode but also the most complex solution. All these methods were cumbersome to implement and troubleshoot. A better solution for integrating IP over Asynchronous Transfer Mode was one of the driving reasons for the invention of Multiprotocol Label Switching. The prerequisites for Multiprotocol Label Switching on Asynchronous Transfer Mode switches were that the Asynchronous Transfer Mode switches had to become more intelligent. The Asynchronous Transfer Mode switches had to run an IP routing protocol and implement a label distribution protocol.

Border Gateway Protocol-Free Core

When the IP network of a service provider must forward traffic, each router must look up the destination IP address of the packet. If the packets are sent to destinations that are external to the service provider network, those external IP prefixes must be present in the routing table of each router. Border Gateway Protocol carries external prefixes, such as the customer prefixes or the Internet prefixes. This means that all routers in the service provider network must run Border Gateway Protocol. Multiprotocol Label Switching, however, enables the forwarding of packets based on a label lookup rather than a lookup of the IP addresses. Multiprotocol Label Switching enables a label to be associated with an egress router rather than with the destination IP address of the packet.

The label is the information attached to the packet that tells every intermediate router to which egress edge router it must be forwarded. The core routers no longer need to have the information to forward the packets based on the destination IP address. Thus, the core routers in the service provider network no longer need to run Border Gateway Protocol. The router at the edge of the Multiprotocol Label Switching network still needs to look at the destination IP address of the packet and hence still needs to run Border Gateway Protocol. Each Border Gateway Protocol prefix on the ingress Multiprotocol Label Switching routers has a Border Gateway Protocol next-hop IP address associated with it. This Border Gateway Protocol next-hop IP address is an IP address of an egress Multiprotocol Label Switching router. The label that is associated with an IP packet is the label that is associated with this Border Gateway Protocol next-hop IP address. Because every core router forwards a packet based on the attached Multiprotocol Label Switching label that is associated with the Border Gateway Protocol next-hop IP address, each Border Gateway Protocol next-hop IP address of an egress Multiprotocol Label Switching router must be known to all core routers. Any interior gateway routing protocol, such as OSPF or ISIS, can accomplish this task.

An Internet service provider (ISP) that has 200 routers in its core network needs to have Border Gateway Protocol running on all 200 routers. If Multiprotocol Label Switching is implemented on the network, only the edge routers—which might be 50 or so routers—need to run Border Gateway Protocol. All routers in the core of the network are now forwarding labeled packets, without doing an IP lookup, so they are now relieved from the burden of running Border Gateway Protocol. Because the full Internet routing table is well above 150,000 routes, not having to run Border Gateway Protocol on all routers is a serious consideration. Routers without the full Internet routing table need a lot less memory. You can run the core routers without the complexity of having to run Border Gateway Protocol on them.