Large-scale distributed systems have become pervasive, underlying virtually all widely used services and machine learning platforms. In tandem, our understanding of how to build scalable, robust, efficient, and secure systems is increasingly well-founded. This class will introduce students to the state of the art in distributed systems research and practice and help them identify important open research problems. A specific emphasis of this iteration of the course would be on systems for machine learning.

Course Staff

Person, Role	Email	Office Hours
Arvind Krishnamurthy, Instructor	arvind@cs	Monday 11:00-12:00 (CSE 592)
Pratyush Patel, Teaching Assistant	pratyush@cs	Friday 13:00-14:00, CSE 5th Floor Breakout

Prerequisites

You should have taken the equivalent of CSE 550 or CSE 551 or CSE 451 or CSE 452. Further, we'll assume basic knowledge of distributed-systems topics, including: remote procedure calls (RPC), two-phase commit, distributed time, serializability, and MapReduce. This can be obtained by taking CSE 550 before CSE 552. A motivated student will be able to pick up these topics on their own. See below for some useful resources.

• Implementing Remote Procedure Calls
• Introduction to Distributed Systems Design
• MapReduce: Simplified Data Processing on Large Clusters

Paper Presentations

Each class will focus on a single topic, with discussion oriented around one or two papers. Starting with Lecture 6, students will present papers for course credit. There are six roles for each paper; you are to sign up for each role once during the quarter and for eight papers overall. Please spread these out - do not sign up for multiple roles for the same paper. If, and only if, there is one person signed up for every role for a particular paper, you should double up. For some slots, there will be two students in each role - please work together on a combined presentation. Each role, with the exception of the design role, will be limited to two slides; for the design role, the presenters can use up to four slides. We will circulate a blank google slide deck for each paper, and the students will populate it with their slides. Each paper will have the following roles:

• Context and prior work: Is there context/prior work needed to understand the paper? If so, explain.
• Design: how does the system work and what are the key ideas behind the design?
• Eval: what are the key results from the evaluation?
• Advocate: advocate for the ideas in the paper.
• Skeptic: advocate against the ideas in the paper.
• Open Questions: what further research is suggested by the paper?

The sign-up sheet will be made available within the first couple of weeks.

Slack

We will use the CSE 552 Slack workspace for course announcements and paper discussions (note that this is different than the UW CSE Slack). You will receive an invite if you have signed up for the course. Reach out to Pratyush if you haven't received an invite or are facing troubles creating an account.

Feedback

We would love to hear from you! You can reach us anonymously via feedback.cs.washington.edu.