ECS 256, Performance Evaluation

Norm Matloff

Fall 2004

1 Outline of Coverage

Topics:

emphasis on the modeling process, not just the mechanics of probability theory
special properties of independence, conditional expectation, generating functions
discrete- and continuous-time Markov chains.
renewal processes
statistical aspects of the modeling process, e.g. confidence intervals for Markov chain parameters and simulation output
discrete-event and Monte Carlo simulation
networks of queues

Here are some sample application areas:

networks, e.g. LANs, Web traffic
OS scheduling issues
data mining

2 Course Materials

There is no textbook. Our materials will consist of my written lecture notes and published research papers.

3 Consultation

My office is in 3053 EUII, Ext. 2-1953. My office hours will be Tuesday and Thursday, 12:30-1:30 pm. I enjoy my office hours very much, and look forward to interacting with you during them. I am also available at other times if you have short questions.

You are welcome and encouraged to send me your questions via e-mail. I read my mail every day, including evenings and weekends.

4 Course Prerequisites

You must have had a calculus-based probability course similar to MAT 131 or STA 131A.

Since the example applications will be to the analysis of various aspects of computer systems, it is assumed you have a basic knowledge of concepts such as caches, virtual memory, timesharing OSs, and so on.

General knowledge of computer science which is typical for beginning graduate students is also assumed, e.g. skill in programming and undergraduate coursework in the analysis of algorithms.

5 Homework

5.1 Amount and Type of Work

There will be approximately four written homework assignments during the quarter. These will consist of mathematical analysis and sometimes some simulation programming.

You are encouraged-but not required-do your work with one (1) partner, and submit the work jointly.

Some of the homework will involve computer simulation. However, this will be light, and you will not be spending a major portion of your time on programming. The programming will be done in Python. You are NOT assumed to have prior background in Python; you will learn it during the course. You'll discover that it is extremely easy to learn, and a delight to program in.

You are welcome to seek all the help you need to insure that you have done the homework correctly before you submit it. Also, a homework problem will typically have some kind of checking mechanism.

5.2 Composing and Submitting Homework

Homework is due by 11:59 p.m. on the due date. If you are working with a partner, submit just one copy of your work, with both names on it.

You are required to write up your homework in LaTeX. You are not assumed to have prior background in it, but will learn it as you go through the course. Again, it is very easy to learn.

LaTeX is a major word processing system in computer science, electrical engineering and the physical sciences. It is the main system used by many CS conferences and journals. So, in addition to it being the required form of writing for our class, it will be a skill which will be of general value to you.

LaTeX is quite easy to learn; just go to my tutorial page, at http://heather.cs.ucdavis.edu/~matloff/latex.html. After an initial small example, the best way to learn is to look at sample LaTeX source files; note that all of our course files in http://heather.cs.ucdavis.edu/~matloff/256/PerfModeling have the LaTeX source available for you to learn from.

If you are much more comfortable with a GUI interface, you can use LyX and export to LaTeX; again, see the above Web page for a short tutorial.

Use the pdflatex command, directly producing a PDF file, rather than latex and then something else. Note that figures, if any, must be used from within your LaTeX file. Since you will be using pdflatex, that means that your figures must be either PDF or JPEG files.

You submit the homework by e-mailing me a LaTeX source file and its PDF output, and also your Python source files, if any. It is required that you package them in a .tar file, and submit just that file.

5.3 In-Class Homework Presentations

Each student will be required to present a certain number of homework solutions in class (probably three or four) over the 10-week span of the quarter. During your presentation, I will probably ask you a couple of questions.

If you work with a partner, you must make each of your presentations jointly, i.e. you cannot have one partner present some homework problems and the other partner present others. In each presentation, each partner should be responsible for discussing half of the solution to the given homework problem.

6 Exam

We will have only one exam, given on the last day of lecture, Thursday, December 9. There is no final exam.

The exam is taken on an open-materials basis. Make sure to bring your written course materials, and you are welcome to bring in any other materials you consider helpful, e.g. probability textbooks, English dictionaries, etc.

7 Grading

The grade will be 90% based on written homework and student in-class presentations of the homework solutions.

The remaining 10% of the grade will be based on an exam given on the last day of lecture. Note that a major use of the exam will be in my subsequent writing of letters of recommendation.

8 Class Newsgroup

It is required that you read our class newsgroup, ucd.class.ecs256, every day. This is where homework assignments will be announced, and various other vital pieces of information will be promulgated.

File translated from T_EX by T_TH, version 3.30.
On 29 Sep 2004, 21:20.