Multi-Protocol Label Switching (MPLS) is a technology for the data transport, which was designed originally with two goals: To make the forwarding in routers as fast as in ATM switches, and to supply a mechanism to offer traffic engineering in the core.
MPLS uses labels on the packets to make forwarding decisions in each router. Once one introduces a packet to a MPLS domain, an entrance label is assigned, depending on its destination and it is sent to the next MPLS hop. The next device analyzes the entrance label and, according to its MPLS labels table, it obtains an interface and an exit label, and it sends the packet to the next hop. This way, the devices switch the traffic through the MPLS domain to the exit node. The association of labels creates an unidirectional path, which is called Label Switched Path (LSP).
In theory, when a packet is forwarded and exact match searches are made, instead of longest-prefix-match searches, greater speed in the commutation is achieved. This behavior is no longer significant, since the IP routers and last generation chips allow making longest-prefix-match searches at the same speed. Because of this, MPLS is no longer seen as a mechanism to accelerate the routing procedure. The real benefit of MPLS is that it provides a separation of the routing (control) and the forwarding. This separation allows the application of an only forwarding algorithm (based on labels) that can be used by all the types of traffic and services, like L2VPN, L3VPN, VPLS, NG-MVPN, etc. The MPLS structure, as regards the forwarding, is maintained while the providers evolve by implementing new services and changing the way in which the packets are assigned to each LSP (IP destination, QoS, Multicast group, VPN identifier, etc.). Another important benefit that MPLS has is the ability to implement traffic engineering in the networks, to be able to establish traffic paths that don’t depend on the desitions of the internal routing protocol.
Generalized Multi-Protocol Label Switching (GMPLS) is the new generation of MPLS. In GMPLS, the functionality of MPLS is extended to include a wider range of LSP options to different devices. Traditional MPLS was designed to transport data, using the IP protocol to establish the paths and assign labels arbitrarily (static or dynamically) to each protocol. In MPLS, the Label Switching Routers (LSR) have a forwarding level, which is able to recognize the limits of the packets/cells, and process the headers of the packets/cells.
GMPLS allows the transport of a wider range of technologies, transporting protocols of the physical layer, link layer and network layer. The additional switching techniques support was achieved extending the functions of MPLS. These changes and new additions have an impact on the basic properties of the LSPs (for example, how they communicate and how they reserve the labels, the unidirecctionality and how the errors are spread). In the forwarding level, GMPLS doesn’t recognize packets or cells, and thus, cannot switch data being based on the information transported in the cells or datagrams. These LSR include devices, whose forwarding decisions are based on parameters like time-slot, wave length or physical port. And this behavior divides interfaces of the new LSRs into the following categories:
- Fiber-Switched Capable interface (FSC): Interfaces that switch traffic based on the physical position of the data. For example, we can consider the interface of "Photonic Cross Connect" (PXC) or "Optical Cross Connect" (OXC) equipment, that can operate with one or multiple optic fibers.
- Lambda-Switched Capable interface (LSC): Interfaces that switch traffic based on the wave length on witch the information is received. For example, we can consider equipment like "Photonic Cross Connect" (PXC) or "Optical Cross-Connect" (OXC).
- Time Division Multiplexing Switched Capable interface (TDM): Interfaces that switch traffic based on the time-slot assigned. For example we can consider equipment like "SDH Cross Connect", "Terminal Multiplexer", "Add Drop Multiplexer" and PDH.
- Packet-Switched Capable interface (PSC): They recognize packets/cells and can switch packets based on the information of the header (Traditional MPLS).
A circuit will only be established between, or by means of, interfaces of the same type. A hierarchy can be formed by nesting different LSPs, in the same interface or among different interfaces. This process of nesting can also occur among different types of interfaces (LSC, TDM and PSC). This way, a LSP, which starts and ends in LSC interfaces, can be nested inside an LSP, which starts and ends in TDM interfaces.
The use of GMPLS, raises the amount of devices that can participate in an MPLS environment. Devices like Optical Cross-Connect (OXC) or Add-Drop Multiplexors (ADM), whose application area is the lower layers, can participate in the GMPLS domain, signaling paths and establishing them for the transport of data. GMPLS also allows layer 3 devices to participate in the signaling of circuits, like optical channels, and that can profit from the techniques of paths' protection and reestablishment of MPLS in the circuits. GMPLS has been designed to unite the world of infrastructure and traditional transport with the IP layer. For example, in a future, routers cold make additional requirements of bandwidth to the optical network dynamically. Physical transport circuits (optical transport, for example) will be able to be created dynamically as well.