Further Experience with Teaching Distributed Systems to Undergraduates

Further Experience with Teaching Distributed Systems to Undergraduates

Interested in learning more about this approach to teaching? Visit my courses page to learn more: https://www.gregorykapfhammer.com/teaching/

4ae30d49c8cc07e42d5a871efb9bcfba?s=128

Gregory Kapfhammer

October 19, 2005
Tweet

Transcript

  1. Further Experience with Teaching Distributed Systems to Undergraduates Gregory M.

    Kapfhammer Department of Computer Science Allegheny College http : //cs.allegheny.edu/˜gkapfham/ Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 1/12
  2. The First Day of Class Are you ready? You must

    become comfortable with the theory and practice of the design, implementation, and analysis of distributed systems and algorithms I’ll never go back to a single address space again: students realize that building distributed systems is challenging and fun! Ricart and Agrawala who? distributed mutual what? why is Lamport stamping time? epidemic algorithms? I think I just had a page fault: students must manage a variety of complexity sources Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 2/12
  3. Distributed System Challenges Students need to be familiar with (or

    willing to learn!) material about programming languages, operating systems, networks, theory, and algorithms Nuts and bolts: RPCs in C, RMIs in Java, Java virtual machines, Jini 1.2 and 2.0, JavaSpaces, and the CLASSPATH Include concepts from concurrency such as semaphores and monitors with real implementations in Java What are we measuring? students must be able to conduct experiments to evaluate the relative strengths and weaknesses of algorithm and implementation choices Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 3/12
  4. Course Objectives Explore the principles and paradigms associated with distributed

    systems Principles: communication, naming, distributed scheduling, synchronization, mutual exclusion, consistency, replication, and fault tolerance Paradigms: become very familiar with object-based distributed systems using Jini and JavaSpaces Include a discussion of special topics such as distributed hash tables (DHTs), tuple spaces, and data stream management systems (DSMS) Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 4/12
  5. Instructional Objectives Course difficulty should be worn as a badge

    of honor Create exciting laboratory assignments where the students actually implement interesting systems and measure their performance A “gloves off approach” with a safety net: try to hide some complexity while ensuring that students retain perspective and understand many low-level details Require students to keep laboratory notebooks where they record hypotheses, data, observations, and design choices Read many scholarly and a few popular presss articles Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 5/12
  6. Topic: Remote Communication Discovery, code downloading, parameters, transparency Local, remote,

    and distributed objects Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 6/12
  7. Topic: Distributed Scheduling Unshared state, resource matching, and process migration

    Case studies: Sprite, Condor, Frugal, ComputeFarm for scientific computation and distributed testing Load balancing vs. load sharing and the cost/benefit tradeoff Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 7/12
  8. Special Topic: Tuple Spaces Space clients can write, take, and

    read Entry objects How do we measure performance and/or correctness? Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 8/12
  9. Special Topic: OpenDHT Clients can put and get with Sun

    RPC or XML-RPC How do we measure performance and/or correctness? Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 9/12
  10. Special Topic: Data Streams Continuous streams of data and queries

    Case studies: STREAM, TelegraphCQ, load shedding Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 10/12
  11. Conclusions The design and implementation of distributed systems is challenging,

    fun, and rewarding Current distributed systems textbooks are generally very good (there are still surprises, though!) There is a wealth of well-written literature from good conferences and journals Most students enjoy laboratories with Jini, JavaSpaces and Python (OpenDHT) while RPC programming in C is seen as tedious Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 11/12
  12. Resources Your comments, suggestions, and participation are invited! http :

    //cs.allegheny.edu/˜gkapfham/teach/cs441/ Further Experience with Teaching Distributed Systems to Undergraduates, 9th JCM, October 19-20, 2005 – p. 12/12