CS294-4/CS444A "ROC Seminar" Assignment #1

Due Thursday, 9/27/01

This week (9/27), we are modifying the format of the ROC seminar to encourage some more structured discussion of the material that's been presented in the class to date. You (the class) will be running the discussion, so to help make it effective, we'd like you to do some preparatory work.

In our lectures so far, we've presented some background for recovery-oriented computing and motivated why we believe it is an idea whose time has come, but we haven't talked much about how the ROC vision might be implemented. On the ROC Research Overview web page, you will find a description of several techniques that we believe are important in implementing the ROC vision.

We ask you to do the following to prepare for Thursday's class:

• Carefully read the ROC Research Overview web page
  
  
• Select three of the "ROC Research Areas" listed and think about their implications with respect to ROC, the class readings, the speakers we've heard so far, your own background, and any other information you can bring to bear on the topic. You are encouraged to do additional reading and research if you feel it would be helpful; the talks and papers on the ROC page and on the class reading page may be good starting points for further pointers to related work.
  
  Some possible questions to ponder:
• what are the pros and cons of the proposed technique with respect to dependability? to failure recovery? to performance?
• does the technique fit into the recovery-oriented computing philosophy? will it improve recovery? hinder it?
• is the technique's effectiveness supported or refuted by the readings and the lectures given by the class speakers? in what ways? if refuted, does a variation on the technique make more sense?
• how does the approach embodied by the technique compare to traditional fault-tolerance approaches (like reliable hardware, triple-modular-redundancy, replicated/checked state)? does it address similar problems? is it complementary? do you think it is more or less effective?
• how widely-applicable is the technique? what kinds of Internet services will it work on (if any)? what about more traditional services like financial databases, timesharing systems, or scientific computing? 
• how might the technique be implemented? what are the potential costs of the implementation in terms of hardware/software complexity, human involvement in the system, and performance?
• are there synergies amongst any of the three techniques you're considering for this assignment? will they work together in the same system? will they fight each other?
• etc.


• Be prepared to discuss and present your thoughts and conclusions in small discussion groups and with the assembled class. Unlike in previous weeks, this week you will be supplying the conversation and we will be expecting all class members to participate, so be ready.

You are welcome to choose more than three research areas to ponder. If you wish, you may also do this assignment in groups of two (but be sure that both you and your partner are prepared to talk about each topic on your own). You might also use this assignment as an opportunity to start thinking about project topics structured around the ROC research areas.

If you have any questions about the assignment, please direct them to the course staff at roc-staff@cs.stanford.edu.