nodes, and b) the wavelengths used in each fibre
section of this path. The OXCs connecting the
fibre sections are then configured to direct the
signal entering from a certain port to the right
output port as well as assigning the signal a new
wavelength, if required, so that it can be trans-
mitted to the next OXC down the chosen path.
The process is repeated until the signal has
reached its destination. A large volume (flow) of
packets can thus be directed to their destination
node by means of an optical label, which con-
sists of the optical wavelength and the output
port of the OXC, given a certain input port. By
employing optical technology in this way and
performing a single routing decision for a large
volume of packets, the whole routing process
is carried out much more efficiently and the
throughput of the IP network is dramatically
improved.
At the same time, since MPLS was not origi-
nally created with optical technology in mind,
there are certain aspects of it that either cannot
be realised in optical networks or are difficult to
realise in optical networks – and therefore ought
best to be avoided. As a result, the adaptations
and extensions of MPLS protocols to GMPLS
may not be enough when optimising optical net-
works that implement GMPLS:a change of
thought may be needed in addition as compared
with the implementation of MPLS. This stems
from the fact that routing in optical networks
cannot be dissociated from the physical layer the
way MPLS functions can take place entirely at
higher network layers. MPLS has been described
as Layer 2.5. GMPLS is still performing MPLS
functionality, i.e. in interplay with Layer 3, but
at the same time it is at the doorstep of Layer 1 –
bridging exactly the two. Some of the standard
MPLS processes that are not particularly facili-
tated by optical technology, are listed below:
- Label merging is not straightforward to realise
optically and it may require a smart combina-
tion with electronic techniques. - Label stacking, pushing and popping appears
rather complex and expensive to realise opti-
cally – if at all possible; it will probably
require a smart combination with electronic
techniques. Intense R&D work is taking place
in this area. - Label swapping can be carried out optically
(e.g. by wavelength conversion) but it is
expensive and should be minimised in optical
networks.
Provisioning in optical networks: routing
and signalling
Extensions to MPLS signalling and routing pro-
tocols are being developed by the IETF in order
to include the specific requirements of optical
technology.Routing in optical networks consists of the rout-
ing problem – with or without some form of
constraint – and the wavelength assignment
problem. The two are in the general case not
dissociated.One of the basic rules in DWDM is that two sig-
nals in the same fibre cannot have the same
wavelength. When calculating the best path for a
connection, path length, number of hops, and
bandwidth availability are not the only factors
that need to be taken into account. Wavelength
allocation must be an integral part of the routing
algorithm and this needs to minimise wavelength
conversions in the general case – if these are at
all allowed. The main aspect in routing algo-
rithms under development concerns minimising
congestion of OChs and eliminating violations
of the “unique wavelength” rule in DWDM.
Additionally, physical characteristics of the opti-
cal link – such as signal quality, noise, etc. – may
need to be incorporated in the protocols such
that the right routes are chosen depending upon
class of service (CoS) in order to facilitate traffic
engineering. Optical routing algorithms often
involve some type of constraint based routing.Firstly, neighbour discovery procedures are
required in the network to identify the optical
nodes, their connections, etc. This can be based
on existing protocols such as NDP [7]. Also a
link state update is required, which – as the
name implies – provides an update of the status
of all links in a sub-network. If a fully distrib-
uted approach is chosen for path establishment,
then the link update needs to take place at all
OXCs. An IP link state protocol such as OSPF
may be adapted to carry out this function [7].Route computation is based on the information
revealed by the link state update. In a distributed
implementation the request for path establish-
ment is forwarded to the OXC at the ingress
point that is then responsible for the computation
and the establishment of the path. Protection
and/or restoration routes may be optimised glob-
ally and off-line or locally at the OXC and real-
time [8, 9].Path establishment is again carried out based on
IP models. The MPLS architecture for IP net-
works implements either RSVP or Constraint
Routed LDP (CR-LDP). These existing proto-
cols for establishing label switched paths are
being extended to encompass optical technology.