Advanced Topics in Quality of Service (QoS)
ECE/CSC 791Q, Spring 2004

General Information

Objective

The aim of this course will be to introduce the students to the issues of Quality of Service (QoS) in high-speed networks as well as the notion of traffic models and their impact on performance. The focus is on providing recently-developed techniques for QoS issues in the literature with sound mathematical backgrounds. The course will help the students engage in advanced research areas in the network performance analysis and also help understand and write technical papers in their careers.

Description

Short Description: (for departmental course listing)

QoS issues in high-speed networks. Introduce advanced modeling and analysis techniques including network calculus (max-plus algebra), introduction to large deviation, Gaussian traffic modeling, long-range dependence, as well as probabilistic limiting results in the literature. A brief summary of probability measure and various inequalities are also given.

Full Description

QoS issues in high-speed networks are (and still will be) one of the major challenges in the design of networks, and hence it is critical that students be taught the methodology and models, which form the basis of determining QoS. This course will be methodological, and the principal approaches for calculating and providing QoS will be given. Both the deterministic as well as statistical approaches to QoS will be covered. To this aim, this course will start with some covering of basic probability theory (measure-theoretic). Then, the students will be introduced to tools from max-plus calculus or network calculus, basic techniques from large deviations theory and also from various traffic models including the issues of short-range and long-range dependence. Applications to admission control and statistical multiplexing of many sources in networks will also be discussed. Throughout the course, proofs and intuitive remarks will also be provided to each topic as much as possible. See the tentative course structure below for more details.

Time and Place

Instructor

Prerequisites

•  ECE 514 or equivalent (a course in probability or statistics),

•  ECE/CSC 570, *and* ECE/CSC 579 or equivalent.

•  Other advanced networking courses such as ECE/CSC 776, 777, etc., are helpful, but not required.

Exam

There will be no exam for this course. However, homework assignments and one final project (essay or report on specific subjects discussed in class) will be given for this course.

Course Website

Textbook

References

• Mischa Schwartz: Broadband Integrated Networks, Prentice Hall, 1996
• C-S. Chang: Performance Guarantees in Communication Networks, Springer-Verlag, 1999
• Jean-Yves Le Boudec: Network Calculus Book, Springer-Verlag LNCS 2050, 2002 (This book is available on the Web for free-download by the author)
• Dembo and O. Zeitouni: Large Deviations Techniques and Applications (2^nd), Springer-Verlag, 1998
• Class notes and recent papers from the literature

Note:  The books above are not required, and only parts of them will be used in the class. As to the probability background, some early chapters of the following books will be useful:

• Sheldon Ross: Stochastic Processes (2^nd), John Wiley & Sons, 1996

•  Patrick Billingsley: Probability and Measure (3^rd), Wiley, 1995

• William Feller: An Introduction to Probability Theory and Its Applications, Volume 1, (3^rd), John Wiley & Sons, 1968

• Richard Durrett: Probability: Theory and Example (2^nd), Duxbury Press, 1996

Tentative course structure^*

• Probability measure
• Borel-Cantelli lemma, Convergences
• Law of large numbers and central-limit theorem (also Poisson limit theorem)
• Various and useful inequalities in L^p space (e.g., Jensen, Holder, Chernoff, Hoeffding, Minkowski, etc.)
• Fluid queueing models: stability, Loynes’ theorem and Lindley’s formula
• Deterministic traffic models (traffic envelopes)
• Introduction to network calculus: leaky-bucket descriptions (or regulated traffic), I/O properties, etc.
• End-to-end delay in networks via network calculus
• Gaussian traffic models and network implications
• Introduction to large deviations (Why large deviations?)
• Case study: large-buffer asymptotic, many-sources-asymptotic
• Statistical multiplexing and effective bandwidths, and Other QoS issues.
• Introduction to the convergence of probability measure (weak convergence)

Note*: This is a bit aggressive and extensive schedule. If this turns out to be too much, it will be appropriately adjusted during the semester.

Students with disabilities

Academic integrity