3.14 MPLS, MPλS, AND GMPLS
The generic multiprotocol label switching (GMPLS) protocol is a generalization of
a previously defined IETF multiprotocol wavelength switching (MPλS) in order to
support very high bit rates and lightpath connectivity in optical and, particularly,
WDM networks. MPλS was an extension of the initially defined multiprotocol label
switching (MPLS) protocol to include features needed for optical networking.
According to MPLS, one or more labels are attached to IP packets when they enter
a label edge router (LER) of a MPLS network domain. Labels indicate the next
router destination in the MPLS network* (Figure 3.24).
The basic principle is that the control plane, which deals with routing, is decoupled
from the switching, which deals with packet forwarding. Thus, when a label-
*Labels have been calculated according to a search algorithm and signaling messages that identify and
establish the best path throughout the MPLS network; the path from source to destination is known as a
tunnel.
switched router (LSR) receives an MPLS packet, it forwards it to one of its outputs,
which is selected according to the label value in the packet and the port it was received.†
Thus, the LSR function in the router may swap the label in the packet with
another label if the packet is switched to a different output port. The connections established
with the MPLS protocol are called label-switched paths (LSPs). Routing
protocols determine the LSPs for predefined traffic classes, known as forward equiv-
†This implies that a MPLS packet is buffered, the label is read, and a controller manages the switching
resources within the router.
alence classes (FECs). FECs are specified based on constraints such as QoS parameters,
entry port number, and source (originating address). The LSP is defined by the
label attached to the packet. Labels are distributed in the MPLSW network by a label
distribution protocol (LDP), each MPLS node (LSR router) constructs input–output
mapping tables with which it routes MPLS packets (Figure 3.35). Thus, a path is defined
by a sequence of labels as they have been defined and distributed by LDP.
When a failure or congestion is experienced in a LSP, the MPLS protocol provides
protection by rerouting traffic. This may be accomplished either by preestablished
alternative routes (required for time-critical and high-priority MPLS packets)
or by recalculating another route (this requires calculations and signaling). This
means that labels may change.
When MPLS is over WDM, the optical network control plane needs not only to
find the best route available but also to assign and provision a wavelength path.*
That is, to establish lightpath connectivity over the WDM network. This led to the
MPLS extended version, the multi-protocol label switching (MPλS) protocol.
Each GMPLS node advertises its bandwidth availability and optical resources
(that is, link type, bandwidth, wavelengths, protection type, fiber identifier) to its
neighboring nodes. Conversely, when a node becomes active in the GMPLS optical
network, it undergoes a signaling process by which it advertises its resources but
also requests from its neighbors their own; this is known as neighborhood discovery.
Thus, the GMPLS algorithmic run time is short and as a consequence, provisioning
is fast and so is switch to path protection and restoration. Fast restoration is
particularly important because in WDM networks that carry traffic at many gigabits
per second per wavelength, long restoration traffic implies an enormous amount of
*WDM nodes receive several fibers, each having many wavelengths, with each wavelength carrying
packets. Thus, at a node, from an input to output, either the same wavelength must be assigned, or a
wavelength conversion must be made. This is a complex matter that is still being researched in DWDM
networks.
lost packets. Switch to protection is typically according to one of the well-known
strategies, 1+1 or 1:N.
GMPLS includes port switching, λ-switching and TDM. It employs search algorithms
to find the best path available within a mesh optical network topology, to
provide the necessary signaling messages, to establish end-to-end and also link connectivity,
topology discovery, connection provisioning, link verification, fault isolation
and management, and restoration. Such algorithms are:
- the open shortest path first (OSPF); according to this, the routing tables (labels)
in each LSR are defined and distributed by LDP. - the intermediate system to intermediate system (IS-IS),
- the constraint-based routed label distribution protocol (CR-LDP), and
- the resource reservation set-up protocol with traffic extension (RSVP-TE).
These algorithms find the best path that fits in specific policy constraints (that is,
the best path that meets the traffic requirements) such as required bandwidth, traffic
priority, real-time aspects, deliverability, and others, which are familiar to
SONET/SDH and to ATM traffic types and service levels. They also monitor the
established path for congestion and failures.
GMPLS, like MPLS, appends a calculated label to the packet. This label describes
the physical port, the assigned wavelength, and the fiber. If the lightpath enters
another node and departs from it without change in the established lightpath,
then the label remains the same. However, if there is a change in the lightpath (due
to congestion or traffic balancing), then the node finds a new (best) path according
to an algorithm, it sends downstream a RSVP generalized label request or a label request
message in CR-LDP, it re-calculates a new label and it replaces the old one.
The GMPLS node may be viewed as consisting of two functions: one that interfaces
the client side or the GMPLS aggregator, and one that interfaces the optical
WDM network.
The network interface takes the received traffic from the aggregator and optically
multiplexes it in the WDM signal. This function is similar to optical add–drop
multiplexing with an optical cross-connect. The GMPLS aggregator receives client
signals, aggregates them, forms labels and packets, and hands them to the optical
network interface. The functionality of the GMPLS aggregator (or router manager)
includes functions such as:
- Manage and maintain the database of link states of established LSPs
- Manage and maintain a database of all resources available in the node
- The routing controller that reviews the link states and node resources and
finds the best path based on an algorithm (OSPF) - The connection controller that takes actions to establish the calculated best
path, modify, and tear down - Generator RSVP Path/Resv messages for LSP setup, modification, and tear
down.
GMPLS assumes a semidynamic provisioning network model. That is, once the
best path has been discovered and the lightpath is established, it remains unchanged
for a relatively long period. As such, switching nodes are fast cross-connects
and lightpaths are not established on a “call-by-call” basis or in real time.
Instead, the node finds the best path by sending downstream a RSVP generalized
label request in CR-LDP, over a separate path, requesting link establishment from
the LSR node to the LSR node. When all link requests have been granted, the end-to-
end lightpath and the label mapping at each LSR node are established (Figure
3.25).
Three standards organizations are responsible for the definition of these protocols:
- The Common Control and Management Plane (CCAMP) Working Group of
the Internet Engineering Task Force (IETF) is responsible for GMPLS. IETF
also is responsible for the link management protocol (LMP), IS-IS, RSVPTE,
and CR-LDP. - The Optical Internetworking Forum (OIF) is responsible for the optical userto-
network interface (O-UNI) and the external and internal optical networkto-
network interface (O-NNI) - The International Telecommunications Union—Telecommunications Sector
(ITU-T) is responsible for network architectures (G.8080), data communication
network (G.7712), neighbor discovery (G.7714), routing and topology
discovery (G.7715), and signaling and connection management (G.7713 and
extensions). ITU-T has also issued many more related recommendations.
3.14 MPLS, MPλS, AND GMPLS
The generic multiprotocol label switching (GMPLS) protocol is a generalization of
a previously defined IETF multiprotocol wavelength switching (MPλS) in order to
support very high bit rates and lightpath connectivity in optical and, particularly,
WDM networks. MPλS was an extension of the initially defined multiprotocol label
switching (MPLS) protocol to include features needed for optical networking.
According to MPLS, one or more labels are attached to IP packets when they enter
a label edge router (LER) of a MPLS network domain. Labels indicate the next
router destination in the MPLS network* (Figure 3.24).
The basic principle is that the control plane, which deals with routing, is decoupled
from the switching, which deals with packet forwarding. Thus, when a label-
*Labels have been calculated according to a search algorithm and signaling messages that identify and
establish the best path throughout the MPLS network; the path from source to destination is known as a
tunnel.
switched router (LSR) receives an MPLS packet, it forwards it to one of its outputs,
which is selected according to the label value in the packet and the port it was received.†
Thus, the LSR function in the router may swap...
More >>
