CSE 704 Data Center Computing, Fall 2010

Administrative Information

When: Wed 10am-12pm
Where: 224 Bell
Organizer: Steve Ko
Office: 210 Bell
Office Hours: Wed 12pm-3pm

News

09/30/10 New class policies as explained in class.
1. New review deadline: Monday 11:59pm
2. Late or no reviews
  - 0.5 point for each late submission
  - 1 point for no submission
  - U after 2 points (non-inclusive)
3. Class participation: Please prepare at least one thing to talk about in class
09/08/10 Please submit your first set of reviews by next Tuesday on Pig Latin & Dryad.
09/01/10 No review required for the next week (MapReduce & MapReduce Online), but please do read the papers.
08/31/10 Clarification for projects: there has been some misunderstanding, but you can't use this seminar to meet your master's project requirement. You can only use your project for additional 2 credits within this seminar.

Overview

This seminar reviews the latest innovations in data centers. The main focus is to learn about various subsystems that empower large-scale Web services like Amazon, Facebook, and Google. The course covers 5 categories of subsystems - front-end, processing, storage, networking, and virtualization. We will examine a number of systems in these areas and learn about problems they are solving and fundamental principles that guide their designs.

Course Credits

You can take this seminar for either 1 credit (default) or 3 credits (approval needed). With 1 credit, your responsibility is to read papers, write reviews, and present at least one paper to the class. With 3 credits, you will also design and carry out a research project.

Class Structure

There are four class activities required - reading, writing reviews, participating in discussions, and presentation.

Reading: we will read two papers per week. For those of you who are not used to reading academic papers, I recommend reading "How to Read a Paper" by S. Keshav.
Writing reviews: you will write a review for each paper we discuss. The format of a review is answering a series of questions for each paper:
- What is the problem that the authors are trying to solve? Do you agree that it is a problem worth solving or even an actual problem? Why or why not?
- What are the main ideas and/or what is the approach of their solution? What are the strengths and weaknesses?
- Other comments, e.g., what you like about the paper, questions, possible future directions, alternative approaches to the same problem, other papers that provide different solutions, assumptions that you disagree, etc.
The reviews do not need to be long. Usually a few insightful answers/comments are better. You are required to email the reviews to me by 11:59pm every Tuesday.
Participating in discussions: discussions are an important part of the seminar. You are expected to participate in every class. Lively discussion always helps everyone to understand the subject material better.
Presentation: you will present at least one paper to the class. In order to prepare this, you are expected to schedule a time with me at least a week before your class presentation for a practice run. Use this as an opportunity to improve your presentation skills.

Schedule

Date	Topic	Required Readings	Presenters
Week 1 (9/1)	Intro		Steve (pdf)
Week 2 (9/8)	Processing - 1	MapReduce: Simplefied Data Processing on Large Clusters , J. Dean et al., OSDI 2004 MapReduce Online , T. Condie et al., NSDI 2010	Steve (pdf)
Week 3 (9/15)	Processing - 2	Pig Latin: a Not-So-Foreign Language for Data Processing , C. Olston et al., SIGMOD 2008 Dryad: Distributed Data-Parallel Programs from Sequential Building Blocks , M. Isard et al., EuroSys 2007	Gaurav Harsh
Week 4 (9/22)	Detour	Chord: A Scalable Peer-to-peer Lookup Service for Internet Applications , I. Stoica et al., SIGCOMM 2001 The Design and Implementation of a Log-Structured File System , M. Rosenblum et al., SOSP 1991	Ashwini Kyungho
Week 5 (9/29)	Storage - 1	The Google File System , S. Ghemawat et al., SOSP 2003 The Chubby Lock Service for Loosely-Coupled Distributed Systems , M. Burrows, OSDI 2006	Puneet Sumit
Week 6 (10/6)	Networking - 1	Ethane: Taking Control of the Enterprise , M. Casado et al., SIGCOMM 2007 OpenFlow: Enabling Innovation in Campus Networks , N. McKeown et al., SIGCOMM CCR 2008 (also http://openflowswitch.org) NOX: Towards an Operating System for Networks , N. Gude et al., SIGCOMM CCR 2008 (also http://noxrepo.org)	Sarah Anand Branislav
Week 7 (10/13)	Front-End - 1	Protection and Communication Abstractions for Web Browsers in MashupOS , H. J. Wang et al., SOSP 2007 Secure Web Browsing with the OP Web Browser , C. Grier et al., Oakland 2008	Herat Richa
Week 8 (10/20)	Virtualization	Xen and the Art of Virtualization , P. Barham et al., SOSP 2003 Live Migration of Virtual Machines , C. Clark et al., NSDI 2005	Jangyoung Karthik
Week 9 (10/27)	Storage - 2	Speculative Execution in a Distributed File System , E. B. Nightingale et al., SOSP 2005 Everest: Scaling Down Peak Loads through I/O Off-Loading , D. Narayanan et al., OSDI 2008	Lei Akanksha
Week 10 (11/3)	Front-End - 2	Secure Web Applications via Automatic Partitioning , S. Chong et al., SOSP 2007 Mugshot: Deterministic Capture and Replay for JavaScript Applications , J. Mickens et al., NSDI 2010	Madhubabu Prateek
Week 11 (11/10)	Networking - 2	PortLand: A Scalable Fault-Tolerant Layer 2 Data Center Network Fabric , R. N. Mysore et al., SIGCOMM 2009 VL2: A Scalable and Flexible Data Center Network , A. Greenberg et al., SIGCOMM 2009	Gudi Ashwin
Week 12 (11/17)	Storage - 3	Dynamo: Amazon's Highly Available Key-Value Store , G. DeCandia et al., SOSP 2007 Bigtable: A Distributed Storage System for Structured Data , F. Chang et al., OSDI 2006	Vaibhav Suraj
No Class (11/24)		No Class
Week 13 (12/1)	Networking - 3	Onix: A Distributed Control Platform for Large-scale Production Networks , T. Koponen et al., OSDI 2010 Safe and Effective Fine-grained TCP Retransmissions for Datacenter Communication , V. Vasudevan et al., SIGCOMM 2009	Anil Amol
Week 14 (12/8)	Closing Thoughts		Steve