Each week one student will volunteer to present a paper of his/her choice. We strongly suggest that you pick a paper from a top conference (SOSP, OSDI, ASPLOS, SIGCOMM, etc). The presenting student will be responsible for leading the discussion. See the “Guidelines for presenters” section below for more information.
Guidelines for participants
The purpose of this reading group is to have a useful discussion about the paper at hand. To achieve this, it is important that participants have already read the paper and understand it to a reasonable extent. You don't need to understand every single thing about the paper–feel free to ask questions about the things you don't understand–but you should at least know what the paper does and how.
Please do not just skim through the paper and expect to learn the rest during the reading group meeting. As a rule of thumb, if you have not spent at least 1-2 hours reading the paper, this is a good indication that you should not attend this week's meeting.
Finally, we suggest that you take a look at the next section, “Guidelines for presenters”, even if you are not presenting yourself. Given that the presenter is not supposed to do all the talking, but is mostly there to lead the discussion, it would be useful to take a look at the questions around which the discussion is likely to evolve; keep them in mind while reading the paper.
Guidelines for presenters
The primary role of the presenter is to guide the discussion. Since we assume that everyone present will have read the paper, the presenter is not required to explain what the paper is doing. That said, a quick 1-minute summary might be appropriate to get things started. Also, when possible, it would be very useful if the presenter can give an overview of the related work and how this paper fits in the general picture. Other than that, there are a number of things that would be interesting to talk about:
- Paper strengths: what makes this paper interesting? Can this idea/technique be applied somewhere else?
- Paper weaknesses: are you not convinced it works/matters? Why is that?
- The next step: are there open problems to be addressed?
- Evaluating the evaluation. This is an important issue that is usually overlooked. How convincing was the evaluation? Was there a particular graph missing that you would like to see? Which graph did you think was the most useful/convincing one? Is there some different representation that you think would work better for conveying the experimental results? A useful exercise is to think about what you would like to see in the Evaluation section before you read it.