Instructor: | Peter A. Dinda |
Time: | Winter 2001, Wednesdays and Fridays, 10:30-12 |
Location: | CS Conference Room |
This graduate-level course will focus on how we can measure, analyze and predict the behavior of distributed computing environments. For the most part, we will focus on a statistical signal processing based approach, but we will also touch on queuing theoretic approaches. We will generally read 3-4 papers or equivalent materials each week, covering fundamental ideas and important recent results. Each paper will be presented to the group by a student and then discussed in a round-table manner.
In addition to readings, students will be strongly encouraged to apply what they are learning by using analytical tools such as Matlab and SPlus to study real trace data. Students will also be encouraged to play with on-line measurement and prediction systems such as RPS, Remos, and NWS. Finally, each student will complete a quarter-long project in which they will apply what they learn to an area that interests them. The goal of these investigations will be to produce interesting new research results, perhaps even some that will lead to publications.
Restrictions: CS grad students have no restrictions. Others need permission of instructor (which I am pretty relaxed about giving). There are currently 20 slots, and both graduate and undergraduate sections (395/495).
Prerequisites: It would be helpful, but certainly not essential, for you have some familiarity with probability and statistics, and with signal processing theory. Once I know who is in the class, I will make sure that everyone gets up to speed.
Rumors that we will learn how to predict the stock market have been greatly exaggerated.