


Description

This course will cover the basic tools necessary for performance
evaluation of computer and communication systems (details of specific
topics are given below). It is intended as an introduction to
techniques needed to construct and analyze performance models
that are useful in computer and communication systems design. For
instance, such techniques have been used in the design of the Internet,
OS scheduling policies, parallel and distributed systems, videoondemand
systems, and many more. Although these techniques are useful in other
field (e.g., financial modeling), the course will focus on applications
to computer and communication systems.

Who should take this course?

Any student interested in computer and/or communication systems and
the related performance issues, especially those interested in
operating systems, networking, distributed systems, databases,
and multimedia systems. The intent of the course is to provide
the tools necessary for evaluating designs of such systems as well
as for gaining insights that can be obtained.

Tentative Topics Covered

We will cover the following topics in some detail (as time permits):
 Review of probabilities, random variables, and transforms.
 Introduction to stochastic processes, including Markov chains.
 Baby queueing theory.
 Intermediate queueing theory.
 Markovian models with special structure, including aggregation techniques,
stochastic complementation, matrix geometric structure, and so on.
 Sample Path Analysis.
 Transient analysis.
 Reversibility.
 Queueing networks, including product form networks, mean value
analysis, and so on.
 Simulation.

Text Books and References

There are no required textbooks at this point ... there maybe some
later.
The material will be taken from books and research papers; representative
books include:
 Kleinrock, Queueing Systems, Volume I.
 Ross, Introduction to Probability Models.
 Stewart, Introduction to the Numerical Solution of Markov Chains.
We will not, of course, cover all material in these books;
they are listed just to give you an idea of the topics covered
in this course.

Workload

There will be a number of homework assignments as well as one or two
small projects. The homeworks will be assigned but most of them will
not be graded, i.e., they are for your own good (I will post solutions).
The project(s) will be graded; they will involve the use of tools
such as CSIM, ns2, MATLAB, and so on.
A midterm and a final will be given.
Any schedule conflicts involving exam dates must be reported to
the instructor within one week of the announcement of the exam date.

Grading (Tentative)

 Homeworks/project(s): 20%
 Midterm: 40%
 Final: 40%
The weights are approximate and may change by upto 10%.

