Harvard SEAS logo
CS2640 Modern Storage Systems
Paper Discussion

Paper Discussion Preparations

Half of the this course is paper discussion. Each student is required to present one paper during the semester. While we do not ask everyone to write a formal report on the paper, we do expect you to read the paper and think through it carefully. This applies to everyone in the class, not only the presenter, who will also make a slide to present the paper and lead the discussion.

To help you prepare for the presentation, please consider the following questions:

  • What is the main problem the paper is trying to solve?
  • Why is this problem important?
  • What is the key idea or innovation proposed by the authors?
  • What is the tradeoff or limitation of the proposed solution?
  • Are the experiments and evaluations convincing? Why or why not?
  • How does the proposed solution compare to previous work?
  • What are the strengths and weaknesses of the paper?
  • What are the potential future directions or open questions raised by the paper?

Reflecting on these questions will help you understand the paper more deeply and prepare you for a thoughtful presentation and discussion.

Class Structure

Each class we will discuss two papers, and will be structured as follows:

Instructor gives a brief context.
First presentation.
Open discussion.
Second presentation.
Open discussion.

Effective Paper Reading Strategy

Follow this three-pass approach to efficiently read and understand research papers:

Pass 1 (10–30 min): Triage and Mental Map

Decide whether to invest more time and get the paper's shape.

Do:

  • Read title + abstract, then intro (lightly)
  • Skim section headings, diagrams, and key math blocks
  • Read conclusion and quickly scan references

Outcome:

  • Answer: What problem? What contribution? Rough approach?
  • Decide: stop, bookmark, or continue (not needed for the papers you will read for this class)
💡 Tip: Prioritize figures/tables—they encode methods and results directly

Pass 2 (1–2 hrs): Understand & Evaluate

Understand the "how" and "why" without getting stuck in details.

Identify:

  • Problem definition (inputs/outputs, constraints, models)
  • Key idea (the trick, abstraction)
  • Method/architecture/algorithm (restate-able level)
  • Assumptions (explicit and implicit)
  • Evaluation design (metrics, baselines, datasets, ablations)

Outcome:

  • Write 5–8 sentence summary + strengths/weaknesses
  • Know which parts to revisit for reproduction/extension

Pass 3 (Deep Dive): Re-derive & Stress-test

Be able to use the paper (re-implement, extend, review).

Do:

  • Re-derive key steps; work through proofs/invariants
  • Check experimental reproducibility
  • Audit hidden assumptions and edge cases
  • Compare to relevant prior work

Outcome:

  • Explain what they did, why it works, when it fails, next steps

Additional Resources: