Sponsored by:

Table of Contents:

Definition and Overview

1. Introduction

2. Evolution from MPLS

3. GMPLS Issues and Their Resolutions

4. GMPLS Outstanding Issues

Self-Test

Glossary

PDF of this tutorial

MPLS Background and Operation

MPLS extended the suite of IP protocols to expedite the forwarding scheme used by IP routers. Routers, to date, have used complex and time-consuming route lookups and address matching schemes to determine the next hop for a received packet, primarily by examining the destination address in the header of the packet. MPLS has greatly simplified this operation by basing the forwarding decision on a simple label. Another major feature of MPLS is its ability to place IP traffic on a defined path through the network. This capability was not previously possible with IP traffic. In this way, MPLS provides bandwidth guarantees and other differentiated service features for a specific user application (or flow). Current IP–based MPLS networks are capable of providing advanced services such as bandwidth-based guaranteed service, priority-based bandwidth allocation, and preemption services.

For each specific service, a table for a forwarding equivalence class (FEC) is created to represent a group of flows with the same traffic-engineering requirements. A specific label is then bound to an FEC. At the ingress of an MPLS network, incoming IP packets are examined and assigned a "label" by a label edge router (LER). The labeled packets are then forwarded along an LSP, where each label-switched router (LSR) makes a switching decision based on the packet's label field. An LSR does not need to examine the IP headers of the packets to find an output port (next hop). An LSR simply strips off the existing label and applies a new label for the next hop. The label information base (LIB) provides an outgoing label (to be inserted into the packet) and an outgoing interface (based on an incoming label on an incoming interface).

Signaling to establish a traffic-engineered LSP is done using a label distribution protocol that runs on every MPLS node. There are a number of different label-distribution protocols. The two most popular are RSVP–traffic engineering (RSVP–TE) and CR–LDP. RSVP–TE is an extended version of the original RSVP to piggyback and distribute labels on its messages and to provide traffic-engineering capability. CR–LDP was designed specifically for this purpose. Figure 1 shows the flow of label distribution that is carried out by the downstream LER (in this case LER2) while the LSP flow is the reverse.

Figure 1. Figure 1: An MPLS-Based Network

The MPLS framework includes extensions to existing IP link-state routing protocols. These protocols provide real-time coordination of the current network topology, including attributes of each link. MPLS extensions to OSPF and IS–IS allow nodes to not only exchange information about the network topology, but also resource information and even policy information—for example, IP addresses, available bandwidth, and load-balancing policies. Constraint-based routing algorithms use this information to compute the optimal paths for the LSPs through the network and allow complex traffic-engineering decisions to be made automatically when selecting routes through the network.

MPLS Evolution to GMPLS

Within the past year, the International Engineering Task Force (IETF) has extended the MPLS suite of protocols to include devices that switch in time, wavelength, (e.g., DWDM) and space domains (e.g., OXC) via GMPLS. This allows GMPLS–based networks to find and provision an optimal path based on user traffic requirements for a flow that potentially starts on an IP network, is then transported by SONET, and then is switched through a specific wavelength on a specific physical fiber. Table 1 gives a summary of the GMPLS framework.

The basic challenge for an all-encompassing control protocol is the establishment, maintenance, and management of traffic-engineered paths to allow the data plane to efficiently transport user data from the source to the destination. A user flow starting from its source is likely to travel several network spans–for example, an access or edge network that aggregates the flows from multiple users (e.g., enterprise applications) to feed into a metro network that is SONET–based or ATM–based that itself aggregates multiple flows from various edge networks to feed into a long-haul network that uses lambdas to transport the aggregated flow of multiple metro networks. The reverse path is used to deliver data to its destination.

These networks and the typical devices used are shown in Figure 2.

Figure 2. Dissimilar Networks That Carry End-User Traffic

Summary of the GMPLS Protocol Suite

The evolution of MPLS into GMPLS has extended the signaling (RSVP–TE, CR–LDP) and routing protocols (OSPF–TE, IS–IS–TE). The extensions accommodate the characteristics of TDM/SONET and optical networks.

A new protocol, link-management protocol (LMP), has been introduced to manage and maintain the health of the control and data planes between two neighboring nodes. LMP is an IP-based protocol that includes extensions to RSVP–TE and CR–LDP.

Table 2 summarizes these protocols and the extensions for GMPLS.

The details of each protocol and their enhancements are found in the references at the end of this tutorial.

The protocol stack is shown in Figure 3.

Figure 3. The GMPLS Protocol Stack

Note that the IS–IS–TE routing protocol stack is similar to OSPF–TE with the exception that, instead of IP, connectionless network protocol (CLNP) is used to carry IS–IS–TE's information.