office: (617) 358-1062
fax: (617) 353-6457
To that end, we will cover advanced traffic control mechanisms to deliver QoS in the Internet and in enterprise networks. These traffic control mechanisms operate at different time scales: packet transmission time, round-trip time, session time, hours or day, and weeks or longer time scale. We will discuss access and backbone QoS architectures, such as Integrated Services and Differentiated Services defined for next-generation Internet. The seminar will involve the application of various modeling and performance evaluation techniques (e.g., simulation, queueing theory, dynamic flow theory, control theory).
Topics include: high-speed multiservice networks (e.g., Internet, ATM), classic and emerging multimedia applications (e.g., teleconferencing), real-time transport protocols (e.g., RTP), QoS architectures (e.g., Differentiated Services), traffic characterization/shaping/policing (e.g., token bucket), buffer management and queueing disciplines (e.g., CBQ, WFQ, WRED), signaling (e.g., RSVP), QoS routing and traffic engineering, fast packet classification and switching (e.g., MPLS and optical networking), flow control (e.g., rate-based), QoS measurements (passive, active) and adaptation, QoS mapping (e.g., IP QoS over ATM or wireless), QoS in the presence of mobility, computing performance bounds for admission control (deterministic, stochastic), models and approximations.
See Tentative Reading List and Schedule (subject to change). We will read 2 or 3 papers per week.
Basic networking (CAS CS 555 or equivalent)
Algorithms and data structures (CAS CS 112 or 113 or equivalent)
Working knowledge of probability theory (CAS MA 381 or 581 or equivalent)
Mondays and Tuesdays 1:30-3:00 pm, or by appointment MCS 271.
Textbooks (Not Required)
These textbooks are not required. They can serve as reference to provide the necessary background for the material covered in more detail in the papers (see Reading List, most papers are available on-line). You should be able to find most of these textbooks on reserve in the library.
W. Stallings. High-Speed Networks: TCP/IP and ATM Design Principles.
Mischa Schwartz. Broadband Integrated Networks. Prentice-Hall, 1996.
Franklin Kuo, W. Effelsberg and J.J. Garcia-Luna-Aceves. Multimedia Communications: Protocols and Applications. Prentice-Hall, 1998.
Paul Ferguson and Geoff Huston. Quality of Service. Wiley 1998.
Joel Mambretti and Andrew Schmidt. Next Generation Internet. Wiley, 1999.
There are many other networking textbooks, including:
L. Peterson and B. Davie. Computer Networks: A Systems Approach.
S. Keshav. An Engineering Approach to Computer Networking: ATM Networks, the Internet, and the Telephone Network. Addison-Wesley, 1997.
Uyless Black. ATM: Foundation for Broadband Networks. Prentice-Hall, 1995.
Jean Walrand and Pravin Varaiya. High-Performance Communication Networks. Morgan-Kaufman, 1996.
D. Comer. Internetworking with TCP/IP, Volume 1, Third edition. Prentice-Hall, 1995.
C. Huitema. Routing in the Internet. Prentice-Hall, 1995.
W. R. Stevens. Unix Network Programming. Prentice-Hall, Second edition, 1998.
W. R. Stevens. TCP/IP Illustrated: The Protocols, Volume 1. Addison-Wesley, 1994.
A.S. Tanenbaum. Computer Networks. Prentice-Hall, 1996.
S. Paul. Multicasting on the Internet and its Applications. Kluwer Academic, 1998.
F. Halsall. Data Communications, Computer Networks and Open Systems. Addison Wesley.
W. Stallings. Data and Computer Communications. Prentice Hall.
J. Walrand. Communication Networks: A First Course. Aksen Associates.
D. Comer. Computer Networks and Internets. Prentice Hall.
Course Requirements and Grading Policy
There will be 2 in-class (midterm and final) quizzes to test your basic understanding of the main concepts in papers assigned and discussed in class. These quizzes should be quite easy for students who kept up with assigned readings and paper presentations and discussions.
You are expected to actively participate in the paper discussions, which will be generally led by the instructor during the first part of the course.
You will also give a presentation and lead the discussion on 1 or 2 papers during the second half of the course.
The seminar also requires a research proposal to be written in the format of a "real" proposal -- for this, you will refer to the National Science Foundation (NSF) Grant Proposal Guide (GPG), NSF 00-2. Subject to instructor's approval, a team of 2 students may collaborate in submitting one proposal. You will have to write a pre-proposal by around March 2, 2000. The pre-proposal involves up to 5 pages of project description. The final full proposal is due around April 27, 2000. The full proposal involves up to 15 pages of project description. You may be asked to submit your complete final proposal (except for a budget!) using the on-line NSF FastLane proposal submission system. You will give a proposal presentation during the last 2 weeks. The proposal may be based on your current Masters or PhD projects if related to the subject matter of the seminar, or may serve as a starting point for a Masters or PhD thesis!
In your proposal, your solution approach may involve the modeling of a network system and its performance evaluation using analysis and/or simulation, or the implementation and testing of network services on a network testbed. Thus, to obtain pilot results in support of your proposal, you may need good computer programming skills (e.g., in C, C++ or Java) and working knowledge of operating systems (e.g., Solaris, FreeBSD, Linux). Knowledge of basic queueing theory (as in CAS MA 583) and simulation should also be helpful. Note that CAS CS 670 Performance Analysis taught by Prof. Crovella may be taken concurrently.
Your overall grade will approximately be based on the following policy:
More on Proposal
An important component of this course is the proposal assignment. For this, you must submit two reports and then give a presentation. The first report is a pre-proposal and is due around March 2, 2000. The second and final report is the full proposal and is due around April 27, 2000.You are advised to work immediately and seriously on your proposal. The formal of each report should follow the NSF grant proposal guidelines as mentioned above. The choice of a topic is quite flexible. The topic should, of course, involve QoS Network Applications and Systems, and may be in any of the following areas:
Pre-proposal: After reading about the topic you chose, you should identify a particular problem that you propose to work on and also what your objectives are. For example, you might want to consider a new solution approach or consider/evaluate some aspect(s) overlooked in existing solutions. You should also specify how you plan to achieve your objectives -- whether by analysis, simulation or some other technique. So in your first report, you should precisely define your problem and clearly state the research question(s), and identify related work, your objectives and solution methodology. I will assess the intellectual merit and impact of your proposed work. Keep in mind that your proposal should not be too trivial. You may want to look at the NSF Review Criteria.
Full Proposal: This report must contain a complete description of your proposed work. So, it will include parts of the first report together with some pilot results/findings, which would provide convincing arguments why your proposal is sound -- in real life, why a government agency or industry should fund your proposal! You should obtain these pilot results using the solution methodology you already specified in the first report. For this, you may use one of a number of tools (see below), including simulators, analytical/numerical techniques (queueing theory, dynamic flow theory, worst-case analysis, etc.) and experimental tools (NeVoT, etc.). We should meet to discuss pilot results that are feasible to obtain by the deadline and that would make your proposal as strong as possible.
Presentation: Following the submission of your final report, you will give a presentation in class. The length of this presentation is subject to time constraints and number of students in class. You are free to use an overhead projector or other audio/video tool to present your proposal. Your presentation should be clear and understandable. You should expect questions during your presentation.
Tools: You are free to use whatever tool you need. A popular tool is discrete-event simulation. You can, of course, write your own simulator from scratch. You can also use one of a number of network simulators already available. Most of these network simulators are written in C or C++ and are documented. However, you may need to modify or extend the simulator you choose. Note that due to time constraints, it might not be feasible to make major changes/extensions to obtain your pilot results. Check the documentation for the capabilities and flexibility of each simulator. Again, come see me if you need help on which simulator to use. Following is a list of some simulators: