Course number 67650 – Spring 2011
Experimental Approaches in Computer Science
This course is complementary to
Topics in Performance Evaluation.
This course is more about measurements, including examples of discovering
unknown facts about workloads or the Internet, and generally the use of
The other course places more emphasis on the methodologies of simulation and
analysis, and presents several case studies of performance evaluation.
The plan is to give these two courses in alternate years.
Class location:|| Shprintzak 213, Tuesday 16–18
Teacher:|| Prof. Dror
Questions:|| send email to feit@cs
Exercises:|| submit exercises and get grades and comments using the
David J. Lilja, Measuring Computer Performance,
Cambridge University Press, 2000.
Raj Jain, The Art of Computer Systems Performance Analysis,
- The experimental method
- Data presentation
- Measuring attributes and performance of computer systems
- Workload characterization
- Models of large systems
- Experimental algorithmics
And look at the (tentative) week-by-week plan of the course.
The grade in the course will be determined by your work throughout the course
There will be 12 weekly exercises, worth 7.5 points each.
The exercises are designed to be reasonably short and simple (at least once
you get the hang of it).
They are typically due 6 days after the topic is discussed in class,
so that we can talk about the results the following week.
Please do the exercises in pairs, and use this to discuss what you are doing,
gain a better understanding, and achieve better results.
If you have a good reason for not doing one or more of the exercises
(e.g. you have milu'im for the whole period) you will be excused, and
the grade will be calculated based on the others (but if your partner has
milu'im, only he gets excused!).
There will be no final exam, but there will be a short quiz in the last class
worth 10 points.
The material for the quiz is all the lectures and exercises.
Thus you should come to the lectures and really take part in all the exercises
(don't alternate with your partner!)
The default grade for a very good exercise is 95.
A grade of 100 is reserved for whoever does something special extra,
or shows some special insights.
Note that 100 is not automatic for doing extra work — it has to be
interesting and meaningful.
For more information on various topics, see the
collection of experimental computer science links.
To HUJI CS home page