Statistical Approach to Teletraffic Engineering

Instructors: Prof. Sandrine Vaton*, Ing. Michele Pagano**, Ing. Christian Callegari**

Affiliations:
(*) Ecole Nationale Supérieure des Télécommunication de Bretagne, Département Informatique, Technopôle Brest-Iroise, Brest, France
(**) Dipartimento di Ingegneria dell'Informazione: Elettronica, Informatica, Telecomunicazioni, Università di Pisa

Duration: 18 hours (Module 1) + 12 hours (Module 2)

Period: October 16 - 18, 2006 (Module 1); October 19 - 20, 2006 (Module 2)

Place: Dipartimento di Ingegneria dell'Informazione: Elettronica, Informatica, Telecomunicazioni, via G. Caruso, meeting room, ground floor

Credits: 4 (Module 1) + 3 (Module 2)

Final test: yes

Contacts: Ing. Michele Pagano

This series of lectures is organised in collaboration with the Network of Excellence EuroNGI - Design and Engineering of the Next Generation Internet, within the framework of the PH.D. Programme 2006.

Aims

The goal of this series of lectures is to provide excellent education on IP traffic modelling and data analysis in telecommunication, providing a broad overview of well-known approaches and summarising recent scientific developments. In more detail, Module 1 will review basic mathematical tools in the framework of traffic modelling, and Module 2 will focus on the application of statistical tools to different key topics in networking, such as traffic modelling, performance evaluation, traffic classification and anomaly detection.

Prerequisites

Basics of probability

Syllabus

Module 1: Theorethical Background

• Overview on Markov Chains (6 hours)
• Discrete Time Markov Chains
• Continuous Time Markov Chains
• Embedded Discrete Time Markov Chain
• Markov queues
• Stochastic Models for Teletraffic (9 hours)
• Markovian traffic models: MAP, BMAP, MMPP
• Fitting techniques for MAP and BMAP
• Heavy Tailed distribution
• Long Range Dependence and Self-Similarity
• Wavelet analysis of traffic data
• Multifractals
• EM algorithm (3 hours)
• Estimation of parameters of missing data models
• Forward Bacward procedure for the E step

Module 2: Applications

• Large Deviation Theory (LDT) and Rare Event Simulation Techniques (6 hours)
• Basic LDT results for IID RVs
• Large Deviation Principle
• Application of LDT to queueing systems
• Effective Bandwidth
• Overview on rare event simulation
• RESTART
• Importance Sampling
• Origin Destination Traffic Matrix (3 hours)
• Mathematical definition of the problem
• Discussion on different approaches proposed in the literature
• Statistical Traffic Classification and Intrusion Detection System (3 hours)
• Traffic descriptors
• Classification methods
• Taxonomy and state of the art
• Statistical IDSs
• Markovian models of TCP connections
• Stream data mining