|Politecnico di Torino|
|Academic Year 2016/17|
Network modelling and simulation
Master of science-level of the Bologna process in Communications And Computer Networks Engineering - Torino
The course has the following goals: providing students with the basic tools for the design, dimensioning and performance analysis of communication networks, including theory of elementary queuing networks and their application to communication network modeling. Additionally, it aims at introducing the principles and basic techniques of simulation, focusing on the use of simulation for the design, dimensioning and performance evaluation of a communication network. During the course, students will be introduced to simulation tools and specifically to OMNET++, one of the main OpenSource network simulators; through it, they will simulate relevant case studies of Internet-based protocols and services.
Expected learning outcomes
Knowledge on the basic theory of those stochastic processes that are useful to describe and solve problems (with special attention to applications in the fields of digital telephony and of the Internet) characterized by non-deterministic time evolutions and unknown behavior of lifetime of items or systems.
Knowledge of basic structural notions of a simulator and, specifically, of a network simulator
Knowledge of the characteristics of various types of simulators
Knowledge of potentialities and limitations of simulators for the investigation of computer networks
Knowledge of the importance of tuning the duration of simulation runs with respect to the system at hand and the expected results
Knowledge of the OMNET++ simulator through either in-class demonstrations or individual use aimed at crafting simulation projects.
Ability to build and solve models of communication networks using queuing theory.
Ability to describe and solve simple probabilistic models to study delays in queues, as well as network performance or software reliability problems, and to solve them with both analytical and numerical methods.
Ability to make critical decisions on what aspects of a system should be modeled.
Ability to understand which of the presented processes is more appropriate in the analysis to be performed.
Ability to understand the meaning of the values assumed by the mathematical objects presented during the lectures.
Ability to understand and discuss numerical results.Ability to autonomously set up complex network simulations using native OMNET++ modules,
Ability to develop new OMNET++ modules.
Ability to analyze simulation outputs and gauge their reliability.
Prerequisites / Assumed knowledge
Students should have basic knowledge of probability and stochastic processes, with a focus on continuous and discrete Markov chains. Furthermore, they are expected to know the C/C++ programming languages and be familiar with probability theory. Knowledge of communication network architectures, especially those based on the TCP/IP protocol stack, is also required.
The course is divided into two halves, each 60-hour long. The first half addresses the following topics:
Introduction to queues and general concepts (4 hours)
Markovian queues (M/M/1, M/M/c, M/M/c/0, etc) (12 hours of classes plus 6 hours of examples and problems)
Littles formula, Erlang and Engset formulas (6 plus 2)
M/G/1, G/M/c queues (4 plus 2)
Markovian queuing networks (10 plus 6)
Markovian models of wired and wireless networks (8 hours)
The second half addresses the following topics:
Introduction to simulation techniques (4h)
Classification of simulators (8h)
o Discrete-event simulators
o Message-passing simulators
Pseudo-random number generators (4h)
o Techniques to generate pseudo-random variables
o Statistical tests on pseudo-number generators
Transient analysis and confidence intervals (4h)
The OMNET++ simulator: architectures and simulation description syntax (20h, including 10h in lab)
Modelling and simulating communication networks (20h, including 10h in lab)
o Basic network protocols (Aloha, CSMA)
o Local Area Network simulations
o TCP congestion control simulations
The first half of the course includes lectures and exercise sessions in classroom. Lectures will be devoted to theory, while exercise sessions will be devoted to problem solutions. The second half of the course includes 20 hours in lab. Students will be divided into small groups and asked to developed simulation projects.
Texts, readings, handouts and other learning resources
The teaching material (slides of the lectures, exercises and examples of written exams, both with solutions) will be made available by the class teachers on the didattica web portal.
The following books are useful as a reference:
Sheldon N. Ross, Stochastic processes, Ed. John Wiley, any edition.
Probability, Markov Chains, Queues, and Simulation: The Mathematical Basis of Performance Modeling, by William J. Stewart.
J. Banks, J.S. Carson, B. Nelson, D. Nicol, Discrete-Event System Simulation, Prentice Hall
OMNET++ User Manual available online on www.omnetpp.org
Assessment and grading criteria
The final exam consists of two separate tests:
the first test calls for the solution of few problems about network modelling, similar in nature to those discussed in exercise sessions (50% of final grade)
The second one will concern be a written test with open questions on the theory and practice of simulation (30% of final grade) and the discussion of a written report on the group simulation project (20% of final grade)
Programma definitivo per l'A.A.2016/17