1. Goals
2. Contents of the Course
   2.1 What is simulation?
   2.2 Discrete event simulation (structure, advantages, limitations)
   2.3 Selection of metrics
   2.4 Modelling load, movement, and propagation
   2.5 Selection of simulation components
   2.6 Basic statistics theory and data analysis
   2.7 Interpretation of simulation results and common mistakes
3. Exercises
4. Project
5. Time Schedule & Timetable
6. References
   
   

1 Goals

The goals of this course are to:

• understand common simulation techniques.
• learn how to plan, set up, and run simulation.
• interpret simulation results properly.
   

2 Contents of the Course

To achieve the goals above the following contents are addressed in the lectures.

2.1 What is simulation?

Simulation is introduced as one of three approaches to the performance analysis of a physical system. All three approaches, i.e. mathematical analysis, measurements in real systems and simulation are dened and their advantages and disadvantages are discussed. It is shown that in the area of communication networks, simulation provides a good trade-o between the advantages of mathematical analysis and measurements and is therefore employed quite often in practice.

A definition of simulation as well as its purpose are discussed, answering the question "What is simulation?" in the context of communication networks. The main aspects of simulation of physical systems are presented and different ways of realizing them ("simulation methods") are discussed.

At first , time-continuous simulation is presented. Afterwards time-discrete simulation is discussed, with the event being the central element of both subtypes, namely event-based and time-based simulation. The advantages and disadvantages of the simulation methods are shown, with the conclusion that in the area of communication networks there are good reasons for employing
time-discrete event-based simulation (in short: discrete event simulation).

2.2 Discrete event simulation (structure, advantages, limitations)

The main structure and components of discrete event simulation (event, system state, system clock and event scheduler) are defined and their operation is shown both intuitively and formally. After a general knowledge of discrete event simulation has been reached, common drawbacks and their solutions are discussed. Those include e.g. methods to minimize the impact of simultaneous events and mechanisms to cope with high run-time complexity, in particular parallel and distributed simulation. Furthermore, limitations of application of discrete event simulation will be discussed.

2.3 Selection of metrics

When conducting simulation it is important to identify the aim of the performance evaluation. Depending on this the modelling has to be done and a set of performance criteria and metrics has to be chosen.

Commonly used performance metrics are presented and common mistakes while
using them are exposed. An important aspect that should also be examined is the choice of simulation time. This aspect also contains the identification of transient state and within this the so called transient removal.

2.4 Modelling load, movement, and propagation

The results of a performance analysis highly depend on an accurate choice of adequate models. Thus, modelling different attributes of the reality like load, movement, and propagation is one crucial challenge.

Common and new traffic models, mobility models, and propagation models are
presented. Furthermore, the pros and cons of the different models are explained as well as implicit assumptions when using them. An important aspect that should not be neglected is the generation of random variates.

2.5 Selection of simulation components

The models mentioned in the previous section have to be included in the simulator. They are one part of the components that have to be selected for simulation. Apart from this, several other models e.g. for the different protocol layers have to be chosen. Existing protocol implementation for the different layer are presented. The existing models of the network simulator ns-2 that is
used in the exercises are of particular interest.

2.6 Basic statistics theory and data analysis

Planning and running simulations does only half the job. Once the simulator has returned log-files and figures as results it is important to extract the right set of information from the regularly large amount of data. As simulation does not provide deterministic "yes" and "no", answers questions can often only be answered with a certain probability. This part introduces basic statistical concepts required for data analysis, for estimating the signicance of a result, and for the comparison of data sets.

2.7 Interpretation of simulation results and common mistakes

Simulation is only a means for solving problems and answering questions. Thus, having statistical figures is not enough, but conclusions have to be drawn from those. Only by interpreting the results correctly, the information that is needed to identify problems and to clarify certain behaviours can be obtained. In order to determine the outcome and success of a simulation, certain aspects (e.g.
variability) have to be kept in mind and the knowledge of common mistakes reduces the risk of misinterpretation.

The last step of an interpretation is often the presentation of the conclusions. General presentation techniques and concepts for data representation help to convey the results.

3 Exercises

The aim of the course is to generate and enhance deeper practical knowledge for simulation in computer networks. Therefore, the contents of the lectures mentioned above are complemented by exercises. The exercises will contain theoretical as well as practical tasks.

The aim of the theoretical tasks is to incite discussing the contents of the lectures. The practical tasks are performed in the lab. The aim of these is that each student learns practically how build models and evaluate the performance of computer networks by using simulation.

4 Project

The project will address selected aspects of an up-to-date research topic. Performance evaluation using simulation will be done in order to achieve results concerning the selected aspects.

Possible topics are:

• performance evaluation of routing approaches in mobile multi-hop ad hoc networks
• evaluating the impact of attacks on multi-hop ad hoc networks
• others - to be discussed.

The aim and approach are discussed within the course. The course will be subdivided in teams. Each team will work on one specified aspect.

5 Time Schedule & Timetable

The course will take place from 21.08.2006 to 01.09.2006.

During the first week there will be lectures from 9:00 to 12:00 and exercises/lab-work from 13:00 to 16:00.

The second week is planned as a project week with the following milestones:

• Monday noon: aim, approach and teams are fixed.
• Wednesday noon: discussion of first results and further approach.
• Friday afternoon: presentation of final results.

During the project we will work in the lab for the whole day.

6 References

[1] Raj Jain, The Art of Computer Systems Performance Analysis, John Wiley & Sons, 1991, ISBN: 0471503363.

[2] Averill M. Law, W. David Kelton, Simulation Modeling and Analysis, Third Edition, McGraw-Hill, 2000, ISBN: 0070592926.

[3] Sheldon M. Ross, Introduction to Probability and Statistics for Engineers and Scientists, Third Edition, Elsevier Academic Press, 2004, ISBN: 0125980574.

[4] The Network Simulator - ns-2, http://www.isi.edu/nsnam/ns/