My plan for computer science research includes applying a solid
experimental research methodology to multimedia, networks and systems
projects. This document describes my research
methodology, includes a link to the publications from
my research, and lists some general research
"The fundamental principle of science, the definition almost, is this:
the sole test of the validity of any idea is experiment." -- Richard
A survey of reputable scientific journals found that computer
science publications contain far fewer valid, reproducible experiments
than do journals from other science fields [LHP+95]. If we, as computer scientists, are to be
as effective in scientific research as are scientists from such
well-established fields such as physics, biology and chemistry we must
follow the scientific methods that have been developed over the
centuries in the other experimental sciences. These time-proven
methods include the following steps:
- Observe. Study the world. Find problems to solve that
matter by looking at existing systems or develop prototypes of future
- Hypothesize. Formulate scientific questions to ask.
Phrase the questions in such a way that valuable information can be
obtained whether the hypotheses are found to be correct or incorrect.
- Design. Define and build the experimental system. Choose
the independent and dependent variables. Perform pilot tests to
determine the effectiveness of the experimental system design.
- Experiment. Run the experiment under controlled
conditions. Modify the parameters in such a way as to test the
- Analyze. Thoroughly examine the results of the experiment
by applying rigorous statistical analysis on the data. Determine the
validity of the hypotheses.
- Report. Communicate the research results to other
scientists. Use established methods of data presentation. Present
the entire experiment such that the results can be reproduced by other
I incorporate a strong experimental component to all my computer
science research. Although all steps in the experimental method are
important, the experiment is often the most difficult step. I believe
this area is my strength. In following the above steps I have
come up with several empirical research publications.
Systems: design, implementation and performance of
distributed operating systems and applications.
performance of multimedia applications and computer
system support for multimedia.
Analysis: benchmarks and capacity planning.
Filtering: using the opinions of peers to predict
information value to an individual.
quality specifications for multimedia applications and
capacity planning for application quality.
varied internetworking aspects of online games, including
the effects latency and loss, server selection and user
[LHP+95] Paul Lukowicz, Ernst Heinz, Lutz
Prechelt and Walter F. Tichy, "Experimental Evaluation in Computer
Science: A Quantitative Study," Systems and Software, January 1995.