CS 791: Network Protocols: Performance and Optimization

Fall 1999

Instructor:  John Byers

MCS 280
Phone: 617-353-8925
Dept. of Computer Science                                                    
Email: byers@cs.bu.edu
Boston University
Fax:    617-353-6457

Class meeting time:     TR 12:30-2:00, Classroom TBA

NB: Class will not meet on Thursday, 9/2. The first meeting will be on Tuesday, 9/7.

Office Hours:    Tuesday & Thursday, 2:00 - ?

Course Overview:

Today's ubiquitous, global networking infrastructure bears little external resemblance to the Arpanet research prototype from which it evolved. Yet surprisingly, the original design of the Internet has proven remarkably resilient to very rapid growth, its protocols have been able to accomodate a wide variety of applications unforeseen by its original designers, and it has been able to incorporate new and diverse technologies such as satellite and wireless links relatively easily.

In this course, we will focus on the design and evaluation of the network protocols which has facilitated the remarkable development of the Internet from a research perspective. We will begin by quickly reviewing the basic design principles and protocols on which the Internet is based (prerequisite material). Then, we will consider case studies where we focus on challenges raised by a particular application or new technology, and study and evaluate the proposed solutions. Case studies will frequently delve deeply into statistical and theoretical techniques.

The case studies we will consider include: reliable transport over wireless networks, video streaming over best-effort networks, scalable reliable multicast to a collection of heterogeneous receivers, endhost architectures to support multimedia flows, and computation in sensor networks.

Our text will be a collection of recent research papers, but simply comprehending the papers will not contribute much to the overall grade. Students will be required to be specialists on some of the case studies and should expect to challenge and be challenged by the instructor and the course participants on those days. Moreover, students will complete an independent semester-long research project of substantial scope in addition to smaller guided group projects for case study specialists.

The seminar is targeted towards high-intensity graduate students who are already doing networking research, or who want to get started. CS 555 or equivalent is an essential prerequisite for this course, as is a strong background in algorithms, probability and statistics.