CS-204: Advanced Topics in Networks


Michalis Faloutsos

The project will be a non-trivial original piece of work. The end result should
be at the level of a decent workshop paper.  Literature surveys are acceptable. However,
they have to be thorough, synthesize the read material, and offer an interesting perspective
on the state of the art.  Note that a reserach project should first examine the current literature
in the area.
I will offer several ideas for projects but identifying a topic is mainly the student's duty.

See my suggested paper structure and suggestions.
Every case is different, by when you diverge too much from the given structure
you should think twice and probably have a very good reason for it.

NEW: All people must form groups of size two or three. Single projects are not allowed.
             Groups of size three will need my approval.

Here is a list of project samples from other people, other courses, other universities. I append the grades I would have given them.

  1. traceroute measurements grade A-.
    Very good target, substance and interesting results. However, it is written very poorly and unless you are very familiar with the problem it is different to understand. Also, previous work has not been overviewd adequately.
  2. Intrusion Detection grade B/B+.
    There is definitely a lot work and effort involved. But the paper does not arrive to any conclusion. Did we detect intrusion or not?
  3. an A+ project is coming up...


I would prefer to have projects in these areas. Other areas we can discuss, but I
can't guarantee that "anything goes" . This year I will not accept any mobile adhoc projects
since we have other classes that cover those topics.

  1. Internet Measurements (Traffic, Congestions, End-2-end Performance)
  2. Topology of Networks: Analysis, Modeling, Generation of communication networks (Internet, WWW, perr-to-peer)
  3. Multicasting (routing, simulations)


    In more detail, here are some ideas. Note some of these things are current research projects
of my students, if you want something that is not overlapping, you should say it from the
beginning (i.e. if you want to possible extend this as your research area ).

    I will distribute in class a handout about more specific projects.

    Attention: Due to the size of the class, I willl not be able to spend as much time as I would
like with each group. For this, you should try to optimize the amount of time you spend with me.
In meetings, you should be well prepared on issues, to get the most out the meeting.
For example,  before you talk to me about your project you should make a thorough search
on the web (Google does wonders), think thoroughly what you want or can do, etc.

  1. Look at existing Internet data (NLANR, CAIDA, etc).  Think what you can do with that. Try to answer questions: what is the end-2-end performance, how often do I see congestion, how long does it last, how can I describe/characterize/model congestion or delay. Can I identify where is the congestion. CAPACITY:Can I find the capacity of the links, can I find how the capacity of the networks is distributed (where are the large capacities, where are the small capacities).  Similar thing for the utilization of the network: is it over or under utilization? PREDICTIONS: can I predict the growth of the network, the growth of the traffic, the evoolution of congestion? Are there correlations between the delays and congestion between different sites?
  2. What is a realistic model for usage for: a link, a gateway router, a backbone link? How are the members of a multicast group are distributed (random vs clustered)? Are there patterns (periodicity vs randomness vs scale-free)
  3. WWW networks. Analyze, model, search. Can we identify patterns in the way web-pages are connected? The www graph is a directed one. How can I model how can I generate an artificial www-graph?
  4. Optical Networks. Configuring optical networks: establishing "virtual links" by "connecting" wavelengths on physical links:
                            A-----B ----C      =>     A-----B ----C
               The new edge is virtual: some wavelengths of link AB gets transformed to some
             wavelengths of link BC, this way node B does not have to"look up" and delay packets
             on these wavelengths (i.e. B does not have to  look up the ddress of the incoming
            packets convert the incoming packets from their photonic to an electronic version)


    Surveys are acceptable but they will have to be truly exceptional to get full marks.
    They should be thorough, substantial, have a clear focus (not a collection of random
     topics and summaries). A good survey provides overivew, compares, discusses the relative
    merits, provides a new understanding of an obscure topic. A survey does not only
    report things: it adds value in attempting to compare issues, identifying problems,
    proposing possible solutions, indicating open research areas.

    I am ashamed to even talk about this but I have to do it.
    Your work would have to be original and not identical to projects for other courses
     Some overlap is allowed but it should be STATED in the differences highlighted.
    You have to  reference all resources that you use.
     Obvious things: a) you are not supposed to cut and paste an existing survey or use
    someone's work  (ie a simulator) without clearly stating it. b) You can not fake results.
    Bad incidents take a tremendous toll on my i) energy, ii) time, iii) happiness.
    I would hate to have to deal with any such incident, and the doers will also hate it too. So, let's
    avoid things that will make everybody's life miserable.

    NOTE: if something is not clear whether it is allowed or not, ask (colleagues, profs, me).
    Especially newcomers will have to learn the local work ethics.