2 Optical Network
Functionality
Though relatively immature, the technology is
available to enable optical networks that use
DWDM not only as a means to increase capacity
in the fibre, but rather as a means to provide
direct connectivity and intelligent re-configura-
bility in the network. DWDM optical channels
are distinguishable as based upon the wave-
length of the channel. Most optical components
are actually inherently wavelength dependent, a
fact that may add a complication to transmission
systems sometimes but it also provides an
explicit way to identify and route/process a sig-
nal without needing to read its content – trans-
parently. It is this quality of optical signals that
gives optical networking a huge competitive
advantage against other technologies – electron-
ics – namely because it makes it inherently scal-
able in terms of speed as well as bitrate-, format-
and protocol-blind.
Although the field of optical networks is in full
expansion, very little optical network functional-
ity has actually been implemented in the net-
work as yet. DWDM is used to increase the
bandwidth in the fibre and provide point-to-point
connections. Optical add-drop multiplexers
(OADMs) are available and can provide direct
optical connections between end nodes, bypass-
ing intermediate nodes and thus eliminating
unnecessary electronic processing. Yet the
OADMs that are installed today are primarily
fixed wavelength, which means that a predeter-
mined set of wavelengths may be tapped out at a
certain node but no programmable re-configura-
tion is possible. This minimises the potential of
these in a network context exactly because it
makes them inaccessible for the management
system. Optical channels are thus still providing
point-to-point connections in these configura-
tions. Commercial solutions with a management
system that allows point-and-click provisioning
at optical channel level are just emerging, how-
ever, automatic provisioning via signalling
directly from a client (network) is still not avail-
able.
In terms of network topology, optical networks
will consist of a multiple of sub-networks as dic-
tated by administrative geographical and techno-
logical factors. These will need to be intercon-
nected by optical links in an arbitrary topology,
i.e. in a mesh topology as the physical network
is in the general case a mesh network. Mesh
topologies make the best use of the available
bandwidth, facilitate load balancing in the net-
work, as well as provide multiple and short
restoration paths. In order to realise optical mesh
networks, optical cross-connects (OXC) are
required, i.e. programmable switching matrices
with several input and output fibres that can
direct optical channels from any fibre input to
any fibre output [1].OXCs are just emerging – therefore currently
expensive – network elements (NEs) and some
time will be needed before the winning tech-
nologies are identified and they become mature,
widely used systems. OXCs allow switching of
channels to different directions in a mesh net-
work and give full flexibility with regard to
physical path selection within the network, thus
enabling re-configurable optical networks. End-
to-end optical channel (OCh) connections can be
established between two end-nodes by reconfig-
uring the OXCs along a certain path that con-
nects these nodes. “Point-and-click” provision-
ing of OCh’s can be achieved this way and pro-
tection or restoration can be carried out fast in
the optical domain, when required. Bandwidth
can be allocated on-demand or “created” at the
parts of the network where it is required. Note
that an OCh is not necessarily all-optical along
the end-to-end link; neither is it necessarily one
single wavelength along the whole path.With channel speeds at 10 Gbit/s being the state-
of-the-art today, and 40 Gbit/s emerging, a chan-
nel count of over 100 channels per fibre and
transmission distances of over 3000 km without
opto-electronic conversion, it has become eco-
nomical to perform switching/routing functions
in the optical domain – in any case in the core
network. IP routers with multiple optical 10 Gbit
interfaces are the state-of-the-art today. Han-
dling such large volumes of traffic electronically
packet by packet creates huge bottlenecks and
impedes total network throughput. Additionally,
opto-electronic (O/E/O) conversions are expen-
sive especially at such bit-rates and should be
avoided as much as possible. It is estimated that
over two thirds of all traffic arriving at a node
is passing-through traffic in the core network.
Therefore bypassing nodes optically brings sig-
nificant cost savings as well as simplifies and
speeds up network functions.Optics does have its shortcomings: there is no
good optical memory as yet and neither a good
optical buffer. These characteristics or limita-
tions determine the way we regard network
architectures today: processing of information
as such is done electronically. The notion is to
move the electronic processing to the edge of
the network as much as possible and carry out
optical network functions in-between.Ultimately, provided some of the limitations are
overcome, signals may be processed optically. A
step before real optical data processing is optical
packet switching [2, 3] or optical burst switching
where packets or (respectively) streams of pack-
ets are encapsulated in an optical “container”