Side_1_360

that is identifiable by its wavelength. This article

is limited to the shorter term, thus to optical cir-

cuit switched networks, so that we will not refer

to optical packet or optical burst switching here.

However, the network architectures that are dis-

cussed in the following are also applicable to

optical packet switched networks.

3 Network Architectures

With regard to architecture alternatives for IP

over optical networks, the determining aspect is

whether and to what degree the control plane of

the optical network will be integrated with IP or

independent of it. The IP and optical control

planes can in other words be loosely or tightly

coupled in terms of, firstly, the details of the

optical network topology, resources, and routing

information that is revealed to the IP layer, and

secondly, the degree of control IP routers have

on optical network elements and thus the degree

to which they can determine the exact paths

through this optical network. Three architecture

options can be identified from this point of view:

The overlay model:In this architecture option

the optical network has full control over its net-

work resources by means of a fully independent

optical control plane (Figure 1). Communication

with its clients, among these also IP, is done via

a well-defined User-Network Interface (UNI) at

the edge of the network where only signalling

information is exchanged. The client networks

request a connection between two edge nodes,

requesting also certain quality related character-

istics for this connection. These characteristics

do not only regard bandwidth but also e.g. delay,

jitter, degree of protection, etc. The client net-

works have otherwise no control over the exact

routing and priority received within the optical

network.

This “bottom-up” model is very popular with

vendors that have long tradition in optical sys-

tems, heavy optical expertise, and only recent

experience with IP (hardware in any case). IP-

centric vendors have also opted for this model in

their first phase products as this model is feasi-

ble in the short term. Here an “intelligent optical

network” carries out part of the network func-

tionality. Another advantage of this model is that

it is a multi-client solution, which can accommo-

date technologies other than IP that many opera-

tors will need to relate to, at least for a while.

Separating the two control planes implies also

that the two parts may evolve, be adapted, and

be optimised independently, which is a good

future-proof policy. The optical network pro-

vides here a universal platform that is not tied to

one specific protocol but is open to any future

new-comers. This aspect is especially important

at this stage when optical technology is experi-

encing intense growth and is therefore exposed

to large changes. The disadvantage of the over-

lay model, on the other hand, is that it requires

the creation of a new control plane that to an

extent duplicates functionality and may intro-

duce delays – repeating that is the old problem

with layered networks. It can be expected that as

IP gradually displaces alternative technologies,

the overlay architecture will at some point

become an anachronism.

IP-router

optical

subnet

OXC

UNI

UNI

NNI

OCh

Independent intelligent network:

routing, signaling, restoration, traffic engineering

(internal MPλS capability optional)

signaling signaling

Figure 1 The overlay model.
An intelligent optical network
has full control over its
network resources and offers
end-to-end wavelength
services to client networks via
a well-defined User Network
Interface