Summary

You are probably familiar with switching concepts in an IP network. The fundamental problem with routing a packet based on the destination IP address is that every hop along the path has to route the packet. As a result, the forwarding table is based solely on the routing table and your IGP's shortest calculated path. If, for any reason, the forwarding path that is derived through the routing table is congested or is experiencing longer-than-expected delays, you have no choice but to forward the traffic down that path. MPLS-based forwarding and MPLS TE can be used to overcome this limitation.

MPLS forwarding is based on a fixed-length label. The label exchange happens in the control plane in much the same way that routing information is exchanged prior to the forwarding of IP packets. Label exchanges are done through a label distribution protocol such as TDP, LDP, RSVP, or BGP. With the exception of BGP-based label distribution, labels are assigned only to IGP-derived routes.

MPLS has two main modes—frame mode and cell mode. In frame mode, a label header is inserted between the Layer 2 encapsulation and the Layer 3 IP header. Cell mode is for ATM LSRs in which the control plane is MPLS and VPI/VCI values are exchanged between ATM LSRs in place of labels. As soon as the VPI/VCI values are in place, the ATM LSRs behave like normal ATM switches and switch cells.

In the forwarding plane, the ingress LER imposes one or more labels on the data packet. The core devices switch the packet to the egress LER based on the top label only. If the ingress LER imposes a single label, the egress LER usually sees a plain IP packet because of PHP.

IP packets that come into a router can be switched out with a label. When the incoming packet is IP, the FIB table is consulted to make the forwarding decision. If the incoming packet is labeled, the LFIB is consulted. The outgoing packet in this case can be either an MPLS or IP packet.

LDP is the standards-based label distribution protocol. It is the one used most commonly to build an MPLS core. It uses UDP multicast packets to discover directly connected neighbors. As soon as a neighbor is discovered, TCP is used to exchange label bindings. LDP can also be configured between nonadjacent neighbors. This is called directed or targeted LDP sessions. Like most routing protocols, LDP uses TLVs, which makes it flexible. LDP can be configured in conjunction with MPLS TE that is based on RSVP—they coexist.

The remainder of this book talks about MPLS TE. This chapter was provided as a foundation in case you are new to MPLS and as a refresher in case you are already familiar with MPLS basics.