08-16-2017, 09:12 PM
[attachment=15195]
1.1 Simulation
Simulation plays an important role in computer-aided analysis and design of any
system. Simulation technique has often been used to model and design communication
networks. Before we build a system, it is important to evaluate the performance
of various designs. In case of complex systems, it is di cult to build an analytical
model that is mathematically tractable. In such cases, simulation of the system is
often the only technique for performance evaluation.
The simulation of large, complex, high-speed networks presents unique problems,
such as setting up networks of di erent topologies and evaluating their performance
parameters[16]. These problems can be addressed using statistical simulation techniques,
such as importance sampling, distributed computing techniques. Thus, simulation
has been playing an increasingly important role in the modeling and design
of communication networks.
1.1.1 Advantages and drawbacks of simulation
The advantage of a simulator, having modeling constructs closely related to the components
of communication networks, is that development time of a target network
becomes considerably small.
The major drawback of basic network simulators is that they are limited to modeling
only network con gurations allowed by the package's standard building blocks.
Thus, if a communication system has some speci c features which are di cult to
model, they have to be approximated to preserve its behaviour. However, there are
several simulators that allow existing modeling constructs to be modi ed or new
constructs to be created, with increased modeling
exibility.
1.1.2 Types of simulation techniques
In this section we discuss about various simulation techniques available.
Process emulation Every component of the target system, i.e., the computer network
to be studied, is represented by a separate process. Each process executes
code corresponding to the code of its component. An advantage of process emulation
is that all components of the target system can be tested. However, in
order to faithfully capture the dynamics of the target system, it is necessary
to periodically synchronize all the processes. The overhead involved in this
process synchronization becomes a serious disadvantage.
Discrete-event simulation In discrete-event simulation, the target system is modeled
by a set of variables and a set of events[11][17]. Associated with each event
is a routine that can update the set of variables. A simulation of the target
system corresponds to a succession of event occurrences of time, i.e., a sequence
of the form (e1, t1), (e2, t2), . . . , (en, en), where (ei, ti) denotes the
execution of event ei's routine at time ti. The simulator executes a loop, with
one event occurrence simulated in each iteration. It maintains the following
two variables : simulated time, indicating the time of the last event simulated;
and event list, consisting of a set of (e, t) pairs, where e refers to an event and t
refers to the time instant when e is scheduled. In each iteration, the simulator
removes an (e, t) pair with the earliest time from the event list, updates the
simulated time to t, and executes the routine corresponding to event e. The
simulation of event e can schedule new events, i.e., add new (e0 , t0) pair(s) into
the event list.