Geoscience Reference
In-Depth Information
In the past, high speed optical fiber technologies were used mainly in the
long-distance and inter-office networks, while local access networks typically used
twisted pair or coaxial cable. However, the current tendency is to bring optical
fiber closer and closer to the end-users, which led to the concept of
Fiber-To-The-
x (FFTx)
, e.g., Fiber-To-The-Home (FTTH), Fiber-To-The-Building (FFTB), and
Fiber-To-The-Cabinet (FTTC). This shift leads to new optimization problems.
At a high level, a telecommunications network can be seen as a set of equipments
(computers, routers, etc.) connected by links of different capacities (e.g., fiber optic
cables). The capacity of a link often depends of the cable itself (nominal capacity)
but also on the equipments installed at its endpoints (network interfaces). For optical
cables, for example, the nominal capacity is very high and the real limit is the
number of different lightwaves that can be transmitted on the cable. This number
of lightwave is in turn determined by the number of interface cards available at the
endpoints. When designing telecommunications networks, the decisions to optimize
can broadly be categorized as follows.
Equipment location:
Due to the hierarchical structure of telecommunications
networks described above, and given that a single network often uses different
technologies simultaneously, an important question is to locate some particu-
lar pieces of equipment dedicated to provide the interface between different
technologies or network levels. Such equipments include add-drop multiplexers,
concentrators, splitters, regenerators, to name a few. These problems are the main
focus of this chapter, as these are typically related to classical location problems.
Link installation:
A long term planning question is to determine a set of cables
connecting all nodes under some survivability criteria. In this context, the
network is seen as a given set of nodes and a set of possible fiber links that
have to be placed between these nodes to achieve connectivity and survivability
at minimum cost. The long-term horizon considered is such that demand data are
not reliable enough, and only topological aspects are considered. For reviews of
these problems, see Kerivin and Mahjoub (
2005
) and Fortz and Labbé (
2006
).
Network dimensioning and routing decisions:
In the mid-term horizon, given a
forecast of the demand matrix for this period and the current topology of the
network, the problem is to compute how the expected demands will be routed
as well as the necessary capacities of the cables. In some models, the addition
of new edges is allowed. These problems involve, at the same time, survivable
design criteria and routing constraints. At an operational level, the focus is on
routing decisions for demands arriving online. For packet-switched networks,
the decisions are decentralized and each node takes the decision on the next node
to visit in order to reach the packet's final destination. These decisions obey to
some protocol rules, and the network operator control on the routing is indirect,
possibly only by tweaking the protocol parameters (such as the arc metrics used
in shortest-paths routing protocols). An in-depth treatment of these problems can
be found in Pióro and Medhi (
2004
).
