From RAD Lab

CS 294-1 Reliable Adaptive Distributed Systems - Fall 2006

The newly-formed RAD Lab's mission is enabling a single person to develop, assess, deploy and operate (DADO) the next-generation Internet app at large scales--to allow her innovative service to scale to the volume of eBay without building an eBay-sized organization to manage it.

In particular we're interested in investigating the use of statistical machine learning (SML) and visualization as essential ingredients in a prototyping platform that would serve as the basis of these experiments.

Sign up for project presentation timeslot

• 13:00 Jeff, Ted, Kelvin, Keenahn Dynamically Scaling the Datacenter
• 13:25 Igor Distributed Top-k Based Monitoring
• 13:50 Greg, Lilia & Nathan RADTools
• 14:15 Pallavi & Mark Power and Performance of Virtual Machines Slides
• 14:40 Gautam, Daekyeong Global Exceptions
• 15:05 BREAK
• 15:30 Andrew, Jeremy: Partial Failure Detection in Internet Services Using AdaBoost
• 15:55 James, Jesse: StarTrace Enhancements and Demo
• 16:20 Michael, Peter: Dynamic Resource Allocation for Power Efficiency
• 16:45 Zhangxi: Disk and Thermal Emulation in Data Center using RAMP

Course summary

In this course we will read about the state-of-the-art in DADO'ing real large-scale systems using the above techniques, hear from the people in industry and research who are doing it, and develop the platform to enable the RAD Lab's vision. Significant course projects will focus on the use of SML and visualization in the DADO process, other systematic methodologies for the DADO steps, and prototyping parts of the RADS platform or services that will run on it. Projects are open-ended and are expected to lead to a result of potentially publishable quality.

Your tour guides will be Profs. David Patterson and Armando Fox, but we will be joined by other Berkeley faculty and guest speakers from RAD Lab affiliate companies such as Google, Sun and Microsoft.

We are looking for students with a strong background and interest in systems, languages/programming systems, and machine learning/AI. Prereqs are not formal, but a rough guideline is that you should be able to pass the Preliminary Exam in one of those areas and you should be comfortable writing code in more than one language.

Course survey (by Tuesday): http://www.surveymonkey.com/s.asp?u=560132489735

Get account on the X cluster (by Tuesday): https://www.millennium.berkeley.edu/account/ (Use cs294 at cs.berkeley.edu as your sponsor.)

Calendar

Labs:

• Lab 1 (due Friday, Sep 8) <li> Lab 2 (due Sunday, Sep 17) <li> Lab 3 <li> Lab 4 </ul>
  • cheat sheet
  • VMs
  [edit]

  Projects

  • project ideas

  
  

  [edit]

  Sign up for project meetings

  Held on Weds. 10/25 in 645 Soda.

  
  

  • 9:30 Greg Gibeling, Nathan Burkhart, Lilia Gutnik Project
  • 9:50 Igor Ganichev Project
  • 10:10 Michael Armbrust, Peter Bodik
  • 12:30 Andrew Dahl, Jeremy Schiff Project
  • 12:50 Pallavi Joshi, Mark Shlimovich
  • 1:10 Ted Schmidt, Kelvin Lwin, Keenahn Jung, Jeff Chen: Project
  • 1:30 Zhangxi Tan
  • 1:50 James Zhang, Jesse Trutna
  • 2:10 Gautam Altekar, Daekyeong Moon Project
  [edit]

  Student pages

  Students, feel free to add your own pages to this wiki below the RADSClassFall06 namespace (like this link for example). Click here if you need some help on editing. Btw, notice the "Show preview" button on the Edit page. Use the RadsClassUser account; you should have the password from the second lecture.

  ---
  Team 1, Lab 1
  Team 2, Lab 1
  Team 3, Lab1
  Team 4, Lab 1
  Team 5, Lab 1
  Team 6, Lab 1
  Team 7, Lab 1
  

  ---
  Team 1, Lab 2
  Team 2, Lab 2
  Team 3, Lab 2
  Team 5, Lab 2
  Team 6, Lab 2
  Team 7, Lab 2
  

  ---
  Team 1, Lab 3
  Team 2, Lab 3
  Team 3, Lab 3
  Team 5, Lab 3
  Team 6, Lab 3
  Team 7, Lab 3
  Team 8, Lab 3

  ---
  Team 1, Lab 4
  Team 2, Lab 3
  Team 3, Lab 4
  Team 5, Lab 4
  Team 6, Lab 4
  Team 7, Lab 4
  

  Team 8, Lab 4