Seminar on Parallel and Distributed Systems (0368-4184) Fall 2016/2017
Adam Morrison
Wednesday, 16:00-18:00 at Kaplun 118
Overview
Parallel and distributed systems drive modern computing at every scale—from the multicore architecture of the individual machine to the geo-distributed systems hosting cloud computing services. The seminar will cover cutting-edge research on such systems. In each meeting, we'll discuss one paper published at a top systems conference (SOSP/OSDI/NSDI/EuroSys/etc.), which will be presented by one student.Requirements and grading
- Being present in every meeting (unless coordinated ahead)
- Presenting a research paper (70%)
Each student will be assigned one paper to present in a 70-80 minute talk. When you present a paper, you are expected to fully understand it (including material not covered by your slides), be familiar with previous work, and be able to compare the paper with relevant related work.
The goal of your presentation is to make sure we understand the paper too, and to facilitate discussion:- provide background about related work;
- explain technically difficult issues in the paper;
- read the paper critically and point out weaknesses of the work.
Most papers have the authors' slides on the web, and a video of the presentation given at the conference is also often available. You may freely use these materials when preparing your presentation, but keep in mind that conference talks are different from your talk in many ways:
- they are 20-25 minutes long, and so don't go into details;
- the audience might be assumed to have prior knowledge;
- most importantly, they are meant to "sell" the work, whereas we want to discuss its weaknesses too.
Feel free to discuss your presentation with me before you give it.
- Handing in short paper reviews (20%)
To encourage discussion, every student (except the speaker) should email me a short review of the paper discussed in the meeting within one week after the meeting.
The report should be up to one page and should be organized as follows:
- Paper contribution: What has the paper contributed to our knowledge? Your task here is figure out the core ideas that are the essence of the paper and to describe them succinctly.
- Your opinion of the paper: What do you think are the strengths and weaknesses of the work?
- Future directions: What would be interesting directions to build on/extend/improve this work? Does it raise any questions for later research in your mind?
Here are a couple of resources on how to read and evaluate research papers:
- Participating in discussions (10%)
Prerequisites
- Operating Systems (0368-2162)
- Computer Structure (0368-2159)
Schedule
| Date | Speaker | Paper |
|---|---|---|
| Nov 2 | Adam Morrison |
Introduction and logistics |
| Nov 9 | No meeting |
|
| Nov 16 | No meeting |
|
| Nov 23 | Adam Morrison |
Background: distributed transactions |
| Nov 30 | No meeting |
|
| Dec 7 | Alon Resler |
Extracting More Concurrency from Distributed Transactions (OSDI'14) |
| Dec 14 | Adam Morrison |
Background: big data systems |
| Dec 21 | Adi Gilboa |
Discretized Streams: Fault-Tolerant Streaming Computation at Scale (SOSP'13) |
| Dec 28 | Eyal Agiv |
Latency-Tolerant Software Distributed Shared Memory (USENIX ATC'16 best paper) |
| Jan 4 | Adam Morrison |
Background: multi-core |
| Jan 11 | Nir David |
Algorithmic Improvements for Fast Concurrent Cuckoo Hashing (EuroSys'14) |