Wireless Network Design: Optimization Models and Solution ...

282 Marina Aguado, Jasone Astorga, Nerea Toledo and Jon Matias
cal layer behavior takes into account pre-calculated results from a MATLAB [25]
implementation.
12.5.1 Traffic Modeling Application
Accurately modeling traffic in the network helps produce an optimal application
and network design. Different data traffic simulation technology strategies may be
followed. A simulator framework may generate traffic in the network explicitly, simulating
all packets and queues in detail. As a consequence high-fidelity results are
obtained. On the other hand, potentially longer simulation run times are obtained.
Other times, the simulation tool may make use of an analytical strategy, simulating
the data traffic at a more abstract level making use of mathematical models. This
implies faster simulation run times. There are also some simulators that exploit hybrid
solutions making use of a combination of both explicit and analytical modeling
techniques. As an example, in [24], mathematical traffic models are introduced for
the most common application services. One of the traffic models studied is the web
browsing (HTTP) traffic model. Although most measurements of HTTP traffic show
that the large majority of page responses consist of relatively small objects, the distribution
of page sizes is heavy-tailed. That is, infrequent but very large page objects
constitute a significant proportion of overall transmitted bytes.
Most of the simulation frameworks allow the final user to select between a set
of standard application models (e-mail, ftp, http, database, voice, video conferencing,
etc.) and to configure the source (user or group of users), destination (user or
servers), traffic patterns, and traffic representation for each service. Each standard
application also allows specific configuration settings. Some simulation tools make
use of specific development tools that allow the end user to create customized applications.
In [1] the authors describe a train-control application built with OPNET’s ACE
Whiteboard tool. The system is a multi-tier application between the train, several
radio block control centers, and the control center. The different train control routines
(movement authority request, position report, etc.) are modeled in the simulation
tool. When creating this application, it was necessary to define the tiers that
would participate in message exchanges and that later on would be identified as
nodes within the simulation scenario. Figure 12.6 depicts the messages exchanged
between the different tiers. The size and HEX format of these messages as well as
the existent dependency between them, and the logical sequence that they follow
are properly modeled in the simulation. The message processing time in each tier
was obtained from a real trace log. By creating these routines and building this new
application in the simulation tool, it was possible to identify the data traffic profile
for this service. The first graph in Figure 12.7 shows the traffic sent by the train and
generated by the train control application.