Brown CSCI 2690: Cloud and Datacenter Operating Systems

Course Description

All of the large-scale web applications and enterprise applications we interact with daily rely on datacenters for compute, storage, networking and security, commonly via cloud computing interfaces. There are many research questions around how to use the datacenter as a computing platform to architect the “best” applications, all while datacenter hardware is rapidly evolving. This is a graduate-level seminar on cloud and datacenter operating systems. We will study topics such as serverless computing, kernel bypass operating systems, tiered and far memory systems, hardware accelerators, disaggregation and machine learning systems. The goals of the class are to (1) learn about the challenges surrounding cloud and datacenter systems, (2) practice reading, analyzing, and presenting systems research papers on these topics, and (3) perform a semester-long research project and present findings to the class.

Course Structure

The course is structured into the following components:

• Semester-Long Project (40%)
• Reading Responses (30%)
• Class Discussion Lead & Class Participation (30%)
• Warmup Assignment (Optional)

Class Reading, Reading Responses (30%)

For each class, I will assign 1-2 papers for reading and a paper reading response assignment. Readings and responses must be completed before class:

• Responses are due the evening before class, at 4:00 PM ET (posted in the calendar below). You can miss up to three of these summaries with no penalty towards your grade across the whole semester, no questions asked. Note that submitting after 4:00 PM ET is considered a miss. The deadline ensures I have enough time to go through responses to prepare for class.
• Responses will be on Gradescope linked on the calendar below; sometimes they will be a more specific question, while other times I will ask for a paper critique. I will go over how to write a good critique on the first day of class.

Class Discussion Lead & Class Participation (30%)

Attendance is mandatory, as a large part of the class is an in-person discussion. However, it is fine to miss class ocassionally, especially if you are sick or have other extenuating circumstances. Frequent absences (including not having read the reading and not participating in class) will be penalized.

Separately, you will sign up to lead the discussion for at least one class over the entire semester. This specifically involves:

• Sharing a list of discussion points with the instructor at least three weekdays in advance, at 12 PM (Friday for a Tuesday class, or Monday for a Thursday class), to which the instructor will give feedback.
• Leading the class discussion on the day of class, by steering the class towards your list of discussion point.
• You can also optionally meet with the instructor to go over your discussion plan prior to the class (email me to setup a time).

Warmup Assignment (Optional)

Semester-Long Project (40%)

• Projects should be completed ideally in groups of 2-3; if you really would like to work alone, please see me after class, and we can discuss your situation.
• You are allowed to use external research projects for the class project, as long as:
  • The topic of the research has a clear overlap with the topics and themes of the class.
  • You discuss the situation with me beforehand, before the proposal is due.
  • You discuss with your other collaborators on the project and they are ok with you using the research as your class project.
  • You clearly state what work has already happened and what new work you will do over the semester.
  • If the project involves student collaborators external to the class, you must give sufficient evidence to prove you are leading or co-leading the project (e.g., you would either be first or co-first author if submitted for publication) and you must explicitly lay out which parts of the project you are responsible for.
• You are allowed to use the same project as another seminar (2000-level) class, as long as:
  • You clear the situation with me and the instructor of the other class.
  • The topic of the research has clear overlap with the themes of this course.

Project Dates:

Time Breakdown

Reading papers and writing reviews for them takes time. Students should expect to spend 4-5 hours per week preparing for class meetings, in addition to the 3 hours per week of class meetings themselves. This amounts to approximately 80 hours of class work over the semester. An additional 10 will go towards the optional warmup assignment and extra prep towards being the discussion leader. The remainder of class time, approximately 90 hours, is dedicated to the class research project.

Collaboration Policy

You are allowed (and in fact encouraged) to discuss readings, reading questions, and the warmup assignments with others (including people who might not be in this class). However:

• However, all answers you submit must be written by you. Your response should not be copied from anyone else and should not be generated automatically using chatbots.
• For the warmup assignment, while you can discuss your implementations with others, you must write code and responses to the assignment questions yourself.
• You may use AI to help you write code for the research project (and only code, NOT the writeup); however, you must acknoweledge this in the writeup and explain how you used AI (e.g., what tasks you exactly used it for).
• Additionally, you must understand your answers; we reserve the right to interview you about your work to ensure you understand your submissions.
• You must cite all sources (people, websites, papers, etc.) that you consult as a part of your work.

Course Readings Calendar

The readings and questionnaires corresponding to a particular lecture date are to be completed before that date.