In the following you can find information about the workshop TMA 2009 and also information about the keynote speaker. This workshop is sponsored by the IFIP Technical Committee on Communication Systems (TC 6). For TMA 2009 Michel Mandjes from the University of Amsterdam will hold the keynote.

Keynote: Michel Mandjes

Date:

Monday, 11 May, 9:05 - 10:00

Title:

Traffic models, and their use in provisioning and traffic management

Abstract:

Traffic models play a key role in design and control of communication networks. In this talk I will sketch the most important classes of traffic models, and present a number of examples that indicate how these models can be used in link dimensioning and traffic management.

In the first part of my talk, I will review important classes of traffic models, viz. models at the flow level (alpha-fair sharing network, processor sharing, M/G/infty, Erlang loss system), as well as models for traffic aggregates (most notably the Gaussian model, which applies if the aggregation level is sufficiently high, and Markov-modulated inputs). I will (very briefly) summarize the most significant performance evaluation techniques.

The second part of the talk focuses on techniques for link dimensioning, i.e., selecting a (minimal) link capacity such that the users’ performance requirements are met. This obviously requires insight into the interrelationship between the traffic offered (in terms of the mean offered load M, but also its fluctuation around the mean, i.e., ‘burstiness’), the envisioned performance level, and the capacity needed. I first derive, for different performance criteria, theoretical dimensioning formulae that estimate the required capacity C as a function of the input traffic and the performance target, which for the special case of Gaussian input traffic reduce to C = M + A V , where A directly relates to the performance requirement and V reflects the burstiness. As estimating M is relatively straightforward, the remaining open issue concerns the estimation of V. I describe an efficient indirect method for estimating V, which samples the buffer content, estimates the buffer content distribution, and ‘inverts’ this to the variance. Extensive numerical experiments (using a sizable collection of traffic traces from various representative locations) show that the inversion as well as the resulting provisioning procedure work remarkably well.

The last part of my talk consider techniques for detecting load changes in a network link that predominantly carries streaming traffic (e.g., Voice over IP). Modeling the system dynamics by an M/G/infty system, we set up a testing procedure - I'll show how deep probabilistic results (from large-deviation theory) help in designing an efficient test. I then reflect on a couple of approximate variants of lower computational complexity. Simulation experiments are used to assess how well load changes are detected.

About the speaker:

Michel Mandjes received the M.Sc. (in both mathematics and econometrics) and Ph.D. degrees from the Vrije Universiteit (VU), Amsterdam, the Netherlands. After having worked as a member of technical staff at KPN Research (Leidschendam, the Netherlands) and Bell Laboratories/Lucent Technologies (Murray Hill NJ, USA), as a part-time full professor at the University of Twente, and as department head at CWI, Amsterdam, he currently holds a full professorship at the University of Amsterdam, the Netherlands. He was guest professor at Stanford University in 2008.

His research interests include performance analysis of communication networks, queueing theory, Gaussian traffic models, traffic management and control, and pricing in multi-service networks. He is editor of several journals, and has published about 140 papers.

Technical Programme

09:00-09:05 - Opening Address

09:05-10:00 - Keynote Talk

• Traffic models, and their use in provisioning and traffic management
  Michel Mandjes, University of Amsterdam, The Netherlands

10:00-10:30 - Coffee Break

10:30-11:30 - Session 1: QoS Measurement

• Realistic Passive Packet Loss Measurement for High-Speed Networks
  Ales Friedl, Sven Ubik, Alexandros Kapravelos, Michalis Polychronakis, Evangelos P. Markatos
• Inferring Queue State by Measuring Delay in a WiFi Network
  David Malone, Douglas J. Leith, Ian Dangerfield
• Network-Wide Measurements of TCP RTT in 3G
  Peter Romirer, Fabio Ricciato, Robert Franzan, Alessandro Dalconzo

11:30-12:30 - Session 2: Rupture Detection

• Portscan Detection with Sampled NetFlow
  Ignasi Paredes-Oliva, Pere Barlet-Ros, Josep Solé-Pareta
• Automated Detection of Changes in Large-Scale Networks
  Felipe Mata, Javier Aracil, Jose Luis Garcia-Dorado
• Passive Streaming Inference of TCP Connection Structure for Network Server Management
  Jeff Terrell, Kevin Jeffay, F. Donelson Smith

12:30-14:00 - Lunch Break

14:00-15:40 - Session 3: Traffic Classification

• GTVS: Boosting the Collection of Application Traffic Ground Truth
  Marco Canini, Wei Li, Andrew Moore
• TIE: A Community-Oriented Traffic Classification Platform
  Alberto Dainotti, Antonio Pescape
• Revealing the Unknown ADSL Traffic Using Statistical Methods
  Marcin Pietrzyk, Guillaume Urvoy-Keller, Jean-Laurent Costeux
• Accurate Fine-Grained Classification of P2P-TV Applications by Simply Counting Packets
  Silvio Valenti, Dario Rossi, Michela Meo, Marco Mellia, Paola Bermolen
• Detection and Tracking of Skype in a Live 3G Network Exploiting Cross Layer Information
  Philipp Svoboda, Esa Hyytiä, Fabio Ricciato, Markus Rupp

15:40-16:10 - Coffee Break

16:10-17:30 - Session 4: Traffic Analysis & Topology Measurements

• Incentives for BGP Guided IP-Level Topology Discovery
  Benoit Donnet
• Scaling Analysis of Wavelet Quantiles in Network Traffic
  Giada Giorgi, Claudio Narduzzi
• KISS: Stochastic Packet Inspection
  Alessandro Finamore, Marco Mellia, Michela Meo, Dario Rossi
• DTS: A Decentralized Tracing System
  Kenji Masui, Benoit Donnet

17:30 - Conclusion