Fall 2023 - EE 381K

Analysis & Design of Communication Networks

Meets TuTh: 9:30-10:45am ECJ 1.314

Instructor

Gustavo de Veciana
Office: EER 6.874
TuTh 11-12 and Wednesdsay 1:30-2:30
Email: gustavo@ece.utexas.edu
WWW: http://www.ece.utexas.edu/~gustavo

Description

We introduce analytical tools need to construct and analyze models for for communication/computer and other networked systems. The focus of the course will be on discussing tools from queueing theory, optimization, and control, as they apply to evaluating the the performance of various types of systems. The course is intended to build upon an introductory graduate course on probability and stochastic processes and one on communication networks (i.e., EE 381J and EE382N ).

Course Topics

(Will cover as many as I can)

Review Discrete Time Markov Chains: Definition: construction, Markov property, stopping times and strong Markov property Classification of states: recurrence, transience, positive recurrence and null recurrence, periodic. Stationary distributions: Existence, uniqueness and convergence to a stationary distribution. . Positive recurrence and stability: Lyapunov functions to show positive recurrence- Foster's criterion.
Continuous Time Markov Chains and Queues: Counting processes, Poisson processes equivalent definitions and their properties. Nonhomogenous Poisson Point processes; Continuous-Time Markov Chains: global balance equations, time-reversibility, detailed balance, Kelly's Lemma; Examples: birth-death Markov processes, M/M/1 queues, Burke's Theorem, Jackson Networks; Basic queueing: models and notation. Little's Result.
Queueing Network Models for Packet Switched Networks: Kleinrock's Assumptions, routing optimization and optimal capacity allocation
Queues and Queueing Networks: Open queueing networks and Extending Product-form Results: Quasireversible queues, insensitivity and Multiclass random routing; Little's Result revisited: Priority queueing systems, Pollazeck-Khinchin Formula; Closed queueing networks; Truncation of time reversible Markov Chains; Assorted topics: Arrival Theorem, ratio of rates formula, Mean Value analysis and sojourn times distributions in networks.
Loss Network Models for Circuit Switched Networks: Kaufman recursion, large capacity limits; Erlang fixed point approximations; Routing optimization;. Alternative routing and metastability, trunk reservation
Assorted Topics: regenerative simulation; simulated annealing and reversible Markov chains; M/GI/1 queues; Lindley process and stability;
Broadband Networks: large deviations and their use to estimate overflow probabilties; Rate and large buffer multiplexing and effective bandwidths.
Internet Modeling: Resource allocation and utility maximization. max-min fair and other forms of fairness, TCP modeling, mice and elephants Stochastic network models: stability and performance, balanced fair allocations
Stochastic Orderings and Performance Comparisons : Introduce various stochastic orderings, and their use in performance comparisons, e.g., in general do systems with increased arrival/service `variability' see worse performance?
Age of information: Modeling and scheduling for timelineness in systems geared at achieving real-time situational awareness.
Topics in Scheduling: Web Server load Balancing and scheduling: Priority, SRPT, SITA, fat tailed distributions. Wireless Base Station Scheduling: opportunistic scheduling, max weight rules

Prerequisites

This course is intended for graduate students. EE381J Probability and Stochastic Processes or equivalent is a prerequisite for the course. You should also have some background in telecommunication networks, e.g., an undergraduate course on this topic and/or EE382N Communication Networks. If you dont have these prerequisites, you are encouraged to take these first, but may seek permission from the instructor.

Course Web site and Text

I will be using my own notes for this course and will point you to appropriate other resources on the relevant material as I go along. Some pointers to other free texts available on the web can be found at the course web site on the UT Canvas instructional system. A partial set of notes, papers and homework assiginments will be posted as we go along at this web site.

Suggested References

Textbooks below cover much of the material we will discuss!

Stochastic Networks by F. Kelly and E. Yudovina, Cambridge Press, 2014
Communication Networks: An optimization control and stochastic networks perspective by R. Srikant and L. Ying, Cambridge, 2013
Performance Modeling and Design of Computer Systems: Queueing Theory in Action by Mor Harchol-Balter, Cambridge, 2013
Age of Information: A new concept, metric and tool. by A. Kosta, N. Pappas and V. Angelakis, Foundations and Trends in NETworking, Now, 2017

And some old standards!

Multiservice Loss Models for Broadband Telecommunication Networks, by Keith Ross, Springer Verlag, 1995
Reversibility and Stochastic Networks, by F. Kelly, J. Wiley, 1979 (can be downloaded from the web)
High Performance Communication Networks by P. Variaya and J. Walrand, Morgan Kaufman 1996. (Particularly Chapters 6,7)

Grading Policy, Homework, Exam and Project Info

Your work in this class will be evaluated based on the following weights:

35% homeworks
30% single midterm
25% course project
10% class/office hours participation/help with hwk solutions

Homeworks will be the core of this class. They will be assigned on a weekly basis and (typically) due Thursday. You may turn in individual homeworks but are encouraged to work in groups of no more than 4 students and turn in a common homework. It is absolutely critical, however, that you make a substantial effort to understand the details of all the homework. In addition may volunteer to help produce/update solutions to one of the homework sets as part of the course participation -- I will work with you on this and it will contribute to your class participation. I will ask you to submit your homeworks and then subsequently also submit your own self-graded using a grading worksheet.

There is a single midterm exam will be closed, i.e., no cheet sheets, notes or books. It will focus on basics. The midterm is planned to be in class.

For the class project you should find a topic you think might be fun/interesting to you and/or complements your research. You can choose to study and presesnt work that has already been done it the past and/or you can define your own variation or research problem. Be ambitious, the aim here is to explore ideas and concepts. See project section on Canvas for ideas on possible topics.

The project involves several stages

Schedule a chat to informally discuss your topic with me.
Submission of 1 page project description: for format and suggestions see project section on Canvas.
Schedule a followup check-in to let me know how things are going.
In-class oral presentation

The tentative schedule for the exams and project presentations:

Project description: due before Thursday October 5
Midterm 1: Thursday October 5 (tentative)
Project presentation: if possible I would prefer to have some of the presentations early on mixed in with some of the lecture material, but if it does not work out these will happen in the last two weeks of class and everyone's attendence is required.

Where does this course fit in?

In conjunction with this course, or in the sequel you might consider taking: Performance Evaluation, Stochastic Networks, as well as related courses on Digital Signal Processing; and/or Digital Communications and Optimization of Engineering Systems . You might consider taking Wireless Communications; Advanced Signal Processing; and/or Information Theory.

Note: All departmental, college and university regulations concerning drops will be followed. The University of Texas at Austin provides upon request appropriate academic adjustments for qualified students with disabilities. For more information, contact the Office of the Dean of Students at 471-6259, 471-4241 TDD or the College of Engineering Director of Students with Disabilities at 471-4382.