Systems Support for
| RESEARCH PROJECT
Students are expected to engage in a research project for the
duration of this course, and produce a poster presentation, a
research paper and a demo (if applicable) at the end of the course.
Ideally, the research project tackles a current research problem,
and proposes an original solution. The research paper should be
of publication quality.
You are to work in a group of maximum 3 students for the research project.
A good research project must (i) define a problem
(ii) propose a solution (iii) implement the solution (simulated
or real solution), and (iv) evaluate against existing solutions.
Your research project can be of the following natures.
Remember, good research always teaches other researchers something new.
- New research problem/solution You define a new,
interesting problem and propose a solution. Your solution does
not have to be real good (for example, see Bolot and Tullerti's paper
on rate adaptation).
- Existing research problem/new solution You look at
an existing, interesting problem, and propose a new, novel solution
that is better than existing solutions, which can lead to new ways
of looking/understanding the problem (for example, the PIM paper).
- Existing research problem/compare existing solutions
You look at an existing problem and its solutions. Implement
the solutions, compare them and provide new insights to why one
solution is better than another (for example, see Feng and Rexford's
paper on smoothing).
- Build an innovative system Build a novel application
that noone, or few, have built before. This is suppose to be fun and
cool :) But most importantly, identify new issues in your system that
no existing solutions can adequately solve. (for example, see McCanne
and Jacobson paper on vic)
- Empirical analysis of some data collected Researchers
need to build systems that are close to reality. Therefore an in-depth
understanding of what the reality is, is important. For this type of
project, collect some data, perform some analysis and discuss how your
findings affect current research efforts. (for example, see Padhye and
Kurose paper on
client interactions with video server)
I do not expect you to write the code from scratch. There are many
existing libraries/open source applications available, which you may
use for this class. See "software resources" below.
A list of suggested topics are available. However, you are
encourage to propose your own research project, subjected to approval.
16th September Project Proposal Due
13th November Project Poster Session
15th November Research Paper Due
You are required to submit a project proposal by 16th of September. The
proposal should contain at least the following items:
I expect a 1-2 page document. The goal is to (i) challenge you to think
about your project early, and (ii) perform a sanity check on the feasibility
of your idea. The proposal will constitute 5% of your project grade.
- List of team members
- What problem you are trying to solve
- Some idea of how well-studied is the problem (detail list of
bibliography is NOT required)
- Some idea on what is needed to evaluate your solution
(e.g. software and hardware requirements)
- Tentative schedules.
These are just ideas. Each idea can have many open research issues
(some could constitute _a few_ Ph.D. thesis). You should identify a
small problem for your project. Don't try to solve everything :)
- Peer-to-peer Streaming Some file sharing tools allow
users to download from multiple sources simultaneous. Can we stream
from multiple sources, and playback in real-time? Such applications would
allow iTunes like service where customers can listen to an audio (without
ability to save them) before they decide whether to purchase.
- Streaming 3D Graphics Imagine watching "Finding Nemo",
with all the polygons and textures streamed to your desktop. How to
stream these data? how to deal with latency, jitter and packet loss?
We don't stream the rendered images as video so that users can
change their viewpoints dynamically, and different users may have
different viewpoints. (Maybe not so suitable for cinematic entertainment,
but would be nice for games)
- Remote Surveillance Imagine a large group of people
being quarantined for SARS, and someone is monitoring them through
webcam. Or imagine in a battlefield, where many unmanned ariel vihecles
(UAV) are flying around sending video back to the command center. These
applications requires many-to-one communications and have different
requirements than one-to-many communications we learned in class. What
scalability, transmissions, scheduling issues can arise here?
- Device streaming Cellphones and PDAs have limited
computing resources and are battery-powered, and have lossy
wireless network links. What are the challenges in transmission and
playback on these devices?
- Multipath Streaming Traditional network applications
send data through a single path. Recent efforts in path diversity
study how to send data through more than one paths. (say one through
wireless card, another one through IR port/bluetooth/wired link) Streaming
through multiple paths can reduce the effects of bustiness in network
Web Site Design by OoiWT, done with Vim
and Macromedia Dreamweaver MX