Side_1_360

The peer model:In this architecture the control
planes of the optical network and IP are fully
integrated such that IP routers and optical
switching nodes (OXCs) are peers. The two net-
works are merged into a new integrated network
that is managed in a unified way. The optical
network topology is fully visible to routers. A
single protocol is run through all domains and
establishes paths through all network elements
in a seamless manner (Figure 2).

This “top-down” approach is primarily the view
of IP experts and equipment suppliers that have
good expertise in IP but little experience with
optical technology. The view is to keep optical
“dumb and fat” as it has been up to now and rely
on IP intelligence to run the network. The ad-
vantage of this architecture stems exactly from
the fact that it is IP-centric: optimised routes
may be found through the network taking into
account all factors, also physical factors. The
architecture is scalable, functionality is not
duplicated and conflicts between several control
planes do not arise. On the other hand, this
architecture demands that information regarding
the optical network elements is advertised to
routers, resulting in excessive information flows
within the network and an overblown control
system. Thorough adaptations of the routers are
required in that routing information that is spe-
cific to optical networks needs to be incorpo-
rated in the protocols. The degree to which this
creates conflicts is still unclear. Both software
and hardware adaptations can in any case be
required that may not be easy to implement in
the short term. Finally, this architecture is not
inherently multi-client, an aspect that may be
important for some operators (e.g. incumbents).
If the amount of non-IP traffic is relatively
small, this traffic may be carried over IP. How-

ever, leasing of “dark” capacity lines is not facil-

itated by this architecture. Despite its drawbacks,

the peer-to-peer model can be expected to be the

architecture to be adopted in the longer term if

IP indeed dominates the scene.

The augmented model:This architecture can

be a whole range of solutions that lie in the area

between the peer and the overlay model. Here

the network can be seen as comprising separate

IP and optical domains, where each one is using

a separate instance of an interior routing proto-

col. Some reachability information is exchanged

between these domains but the topology of the

optical network is by and large opaque to the

client network. This option may be a good com-

promise in that it is more feasible than the peer

solution and at the same time less rigid and more

efficient than the overlay model. It can be argued

that this solution combines the best of two

worlds because it limits the amount of info

exchange within the network and at the same

time it may allow the delivery of wavelength

services, i.e. trading of pure end-to-end band-

width, that circumvent the IP layer.

The above three architectures were initially seen

as rival solutions. Lately it appears that as the

realities of pros and cons for all options are

becoming more evident, the three solutions are

seen as possible evolution stages down one sin-

gle path. According to this scenario, networks

will first be based on the overlay model, proceed

to an augmented model with enhanced signalling

as well as routing information exchange between

different domains, and finally – assuming IP

becomes the transport protocol – move to the

peer model with a full integration between the

optical and the IP plane. This scenario may be

challenged if good solutions based on the aug-

optical

subnet

OXC

OCh

end-to-end LSP

IP-router

Figure 2 The Peer-to-Peer

model. IP routers and optical

network elements (OXCs) are

peers in a merged network that

is managed seamlessly in a

unified way