COMP_SCI 445: Internet-scale Experimentation



Internet-scale Experimentation is a graduate-level seminar exploring the challenges of large-scale networked system experimentation and measurements. CS 445 Syllabus

  • This course satisfies Project or Technical Elective.
  • Formerly Comp_Sci 397/497 - last offer was Winter 2021

COURSE INSTRUCTOR: Prof. Fabián Bustamante


How would you …

- Evaluate the effectiveness of a feature you added to your startup’ new app?

- Understand the tense relationship between Netflix and ISP?

- Characterize the impact of population growth on urban spaces?

- Understand what determines the quality of experience of Internet users?

- Measure the consequences of network censorship on user experience?

The answers to this seemingly disparate set of questions share a common requirement – carrying experimentation at Internet-scale.

Internet-scale Experimentation is a graduate-level seminar exploring the challenges of large-scale networked system experimentation and measurements. Over the last few decades, networked systems have become an integrated part of everyday life and a critical piece of our economic, educational, health and defense systems. This fact is normally brought up as evidence of the success and broader impact of our field of work.

The other, typically avoided, side of the story is the complications that this translates into for experimentalists. Today it is virtually impossible to run a randomized controlled experiment at even fractions of the scale of many of our systems. Despite this, as we explore new ideas in these uncharted territories we are reasonably asked to provide better evidence of the effects of interventions. In this seminar we will discuss ongoing projects on networked systems experimentation and their applications, in wired and wireless settings, that address some of these challenges.

The class consists of two major components: reading and reviewing papers and doing a research project on your own. For the research part of the course, you will have the chance to work (and expand) some existing platforms and datasets as you formulate and try to answer these and other interesting questions of Internet scale.


  • Introductory notes: Internet architecture, practical issues and good practices for Internet-scale experimentation
  • Experimental platforms: Experimental design, context of experiments
  • Experimental design
  • End-to-end and up-the-stack
  • Network infrastructure
  • Traffic
  • Applications and distributed services: DNS, Web, P2P, VoD, OSN, ..
  • Botnets and other maladies
  • Security and ethical issues

PREREQUISITES: Students must have taken CS 340-0 or CS 345-0 or CS Graduate standing or Consent of Instructor (If you have taken similar courses somewhere else or have not taken any of these courses, but would like to register for this seminar, please contact Prof. Fabián E Bustamante).


The course is organized as a series of paper discussions and a single term-long project.

Most class meetings will be centered around one or two paper presentations and discussion. You should read each paper before coming to class and be prepared to discuss it.

The class will run as mini-conference with you as the Program Committee member. We will use the papers included in the schedule as our set of submissions. Each of you will write reviews for 3-4 of them. We will discuss the papers in a two-part PC meeting (around midterm and the end of the quarter) to decide what paper "should be accepted" for publication.


We will read between 2-4 papers per week. The papers will be first presented to the group by one or more students and then discussed in a round-table manner.

To ensure lively discussions, you will be responsible for reading the assigned papers before each class. I will post a question about each paper 24hr before class. Your answer need only be long enough to demonstrate that you understand the paper; a paragraph or two should be enough. I will check your answers to make sure they make sense and they will count for part of the paper discussion grade.

You may find the following documents useful:


At one time or another, every researcher is asked to review papers submitted for publication at a conference or journal; a process known as peer review. We will work on this skill by running a mini-conference - WINE (We do INternet Experimentation)

All class members will be part of the "Program Committee" for our mini-conference and we will consider all papers listed in our schedule as our submissions.

Each paper will receive three reviews and each PC member will be responsible for writing 3-4 reviews (you are welcome to write additional reviews). We will discuss all papers in a two-part PC meeting (around midterm and during final weeks) to decide what paper "should be accepted" for publication. Each paper discussion will be led by one of the reviewers (assigned by the PC chair).

Some useful readings on reviewing:


Class meetings will be centered around a paper presentation and discussion. Each student will be responsible for presenting one of the papers in the schedule. Giving a good presentation is hard work. Please make sure to allocate enough time to prepare for yours. There are some good pointers around that you may want to look at.

Here is an incomplete list of dos and don'ts:

  • Don't try to present the whole work; remember the talk is just a taster.
  • Think of your primary audience to decide what/what not to expand on.
  • Use examples to motivate the work and approach, and illustrate the key points.
  • Don't put too much on a slide - prune and then prune again.
  • Don't put too much on a slide - just one figure/graph per slide!
  • Don't put too much on a slide - don't waste the header/slide title!
  • Careful with use of animation - not for show, just for clarity
  • Please put numbers in your slides
  • Seriously consider dropping the typical "overview/roadmap" slide
  • Saying enough without saying too much - enough depth to convey your ideas, not so much as to overwhelm your audience


There will be one single project on which you will work throughout the quarter - this is a critical component of the course. Your goal is to propose and tackle a research problem that requires the use of (or explores challenges with) Internet-scale experimentation.

Project reports must be written up in a term paper (due during finals week) and teams will present a summary of their findings at the end of the course as part of a systems class mini-conference. Projects ideas will be suggested by the instructor, but you are strongly encouraged to come up with your own ideas. Based on the topic of your project, you will be assigned a project leader to help you through the quarter.

This is the schedule of meetings and deliverables (mainly to ensure steady progress):

  • Form a group: First week of classes.
  • Project meeting with instructor: Second week of classes.
  • Project initial presentation (you should look at the Questions that any project proposal should answer -- the Heilmeier "Catechism"). The presentation should be 4 slides long, including (1) Project name and team members, (2) What are the research contributions of the project? List of new/interesting concepts to be investigated, (3) Why do we care if you are successful?, and (4) Project milestones and schedule for the rest of the quarter:  Third week of classes.
  • Midterm presentation and report. The presentation should be 4 slides long, including (1) Project name and team members, (2) Revised statement of project goals and list of new/interesting concepts to be investigated, (3) List of issues addressed and pending, and (4) Updated project milestones, highlighting accomplishments to date, and schedule for the rest of the quarter: Fourth/Fifth week of classes.
  • Project meeting with instructor: Ninth week of classes.
  • Final presentation: Finals week.
  • Final report due: During finals week.

The final report has to conform to the format used by the Workshop on Hot Topics in Networks. Reports should be no longer than 6 pages (you can use appendices or a webpage to document details). The following structure is suggested:

  • Abstract: What did you do, why is important and what are your high-level results?
  • Problem statement: What is the problem you tried to solve?
  • Prior work: How others have addressed the problem before and why that was not enough?
  • Research approach: What was your approach to solving the problem? What did you design, build? What was your experimental methodology?
  • Results: What were your results? How did you evaluate your work? What were your figures of merit?
  • Lessons learned and future work: If you knew what you know now, what would you do differently? What questions are left for future work?
  • Summary and conclusions.


In use, will be a criterion-referenced method to assign your grade; in other words, your grade will be based on how well you do relative to predetermined performance levels, instead of in comparison with the rest of the class. Thus, if a test has 100 possible points, anyone with a score of 90 or greater will get an A, those with scores of 80 or greater will get a B, those with scores of 70 or greater will get a C, and so on. Total scores (between 0 and 100) will be determined, roughly, as follows:

  • Paper discussion participation (and questions) 10%
  • Paper review and PC meeting participation 20%
  • Paper presentation 20%
  • Project 50%