CDER Book Project Telecon with Early Adopters

Meeting Minutes: Book Project Telecon with Early Adopters

June 17, 2013 - 2pm

Download Meeting Slides
Download Meeting Audio
Download Meeting Minutes

Minute taker: Michael McDermott, Graduate student, Georgia State University

• Attendees:

  Adams, Joel Ayyash, Moussaka Brown, Richard Baker, Johnni Berlich, Rüdiger Bogaerts, Steven Bunde, David Burkman, Brad Busch, Costas Byrd, Gred Chen, Yong Dechev, Damian Desell, Travis El Amrani, Chaker Elster, Anne C. Giacaman, Nasser

  Gupta, Anshil He, Xidex Iparraguirre, Javier Jha, Sumit Juedes, David Kothapalli, Kishore Lee, Hwajung Li, Zhen Löffler, Frank Lu, Yung-Hsiang Mache, Jens McDermott, Michael Morena Maza, Marc Oh, Tae Ou, SC Ou, Shaochieh

  Pitsianis, Nikos Prasad, Sushil K. Rao, DJ Rivera, Wilson Rosenberg, Arny Scales, Karl Sells, Larry Shamshi, Jawwad Sussman, Alan Syrotiuk, Violet Tanniru, Govardhan Vaidyanathan, Ramachandran Weems, Charles Wilkinson, Barry Yuan, Quan

06172013 - 2pm CEDR teleconference 35 attending, 53% attentive

14:04 - agenda intro and how teleconference is formated

14:06 - CDER intro (4of12)

14:09 - experience and resources

• Kothapalli - DS w Parallel - sometimes its too early for students to appreciate 'why parallel' forced to remove most technical aspects in early classes
• Bogaerts - lack of beginner material, quinn's book helps (cs1) forced to create/adapt content to low level
• Syrotiuk - 4th level class, parallel centric, uses 'parallel studio' (intel) but its complicated tool
• Vaidyanathan - logic design & parallel
• Yuan - moving to new institution, teaching computer organization, andrew tannenbaum - "structured computer organization" : includes on chip parallelism (ch8) these are higher level books
• Lee - pdc - distributed systems by bosch? its really hard to find a good book. on top of this make students do hands on with hadoop
• Brown - tannenbaum book considered more of a lower level book (not high level), ie computer organization. there is a great need for more PDC in textbooks
• Shamshi - teaching 2 courses, not found a single book that covers the depth needed

14:32 - what should we ask participants today? what is missing from the pdc curriculum? (esp in core level)

• Syrotiuk - operating system or systems /should/ be listed (understood to be included by default - prasad)
  • Rosenberg - doesn't like the idea that our curricula tend to compartmentalize its really helpful in the core not to section things off
• Prasad - we really need more input into what is missing/important

14:38 - book project - integrating pdc into core topics
fixing the Lack of suitable books for core topics
part 1 for instructors - what and how to teach, pitfalls, techniques, pedagogy, examples
part 2 for students - supplemental teaching materials, custom textbooks

• Weems - part 2 is harder and takes more care than part 1
• Adams - timeline for book? (yes) it would be helpful if there was a good definition(s) for "distributed"
  1. three different areas that get intermixed. this needs clarification.
  2. distributed systems
  3. distributed multiprocessing (mpi)
  4. distributed computing (seti@home)
• Rosenberg - histortically there has been a much sharper divide between parallel and distributed computing.
• Syrotiuk - what do we mean by 'core' corses
  • Prasad - lower level courses CS1, CS2... etc
• Desell - it seems like it would be difficult to agree on what distributed computing is. As long as the fundamentals are in the book these distinctions may not matter
• Brown - organize into two parts (volumes?) makes sense 1st part - common background directed at instructors makes sense. reaching the broad audience may be difficult for various reasons (school type, goals, etc...)
• Prasad - we have been talking about CS area only, but this does not preclude other disciplines being included
• Rosenberg - challenge: discuss how to teach various concepts in an onion (layered) like structured teach certain PDC topics then revisit in a more advanced course Do instructors feel the need for that?
  • Brown - This is not only useful and interesting to students but it accomplishes good pedagogical goals. helps students emerge with a sense of maturity
  • Syrotiuk - strongly agrees with layered teaching structure. reinforcement learning
  • Byrd - Computer Engineering @ NC State teaching is consistent with this idea of layered teaching
  • Vaidyanathan - students don't realize that when you look at an IC all the things happen on the same clock... pipelining.
  • Desell - is the book going to be designed as modules?
    • Rosenberg - pick a particular crosscutting topic (synchronization) and write an essay that provides one persons perspective on how one could develop a treatment of the topic that has this layered structure. not sure how this would end up as a book. idea: book is a list of essays and each essay is a treatment on a pdc topic that is NOT associated with any particular course my vision of 'core' includes ideas of PDC that every student is exposed to so it comes early in the curriculum. wants people to learn to recognize the essential component of parallelism/concurency in computation
    • Adams - Taking the layered approach ensures that even students outside of the curriculum are exposed to the ideas of pdc
  • Giacaman - what really stimulates the students to take an interest in the topics and engage? active learning techniques illustrate how it relates to the everyday world with examples/analogies. force the student to actively participate in their own education
    • Rosenberg - would love to see an essay on that
  • Bogaerts - essays... modules... here is a different or complimentary approach: one of the strengths of the edupar approach is that early adopters have had the chance to begin seeing/developing how/what actually works. afraid that if the book was only essays that it would be less strongly based on these experiences. we should try and leverage our experiences as much as possible.
    • Prasad - there is no one single approach, we haven't ruled anything out. We want everyone to participate
    • Brown - comment on essay approach - it may be more easily applicable. an essay seems 'comfortable'. it might be less immediately applicable though
      • Vaidyanathan - agrees more or less. the instructor shouldn't have to 'hunt' for the topics for a course.
      • Sussman - it should be up to the instructor to decide if something is applicable to their particular course/institution. we really just want to help people to come in and try this without having the invent everything themselves
  • Amrani - this is still a new topic in morroco, book - "patterns for parallel programming" uses cuda to teach shared memory. students implement a cluster with virtual machines. suggestion : add some hands for students on to the book
  • Adams - 2 approaches being proposed modules vs essays so going back to the two part book structure:
    part1 - essays
    part2 - modules
    • Rosenberg - chapters made him cringe, does not want to compete with some very excellent books that just lack in the PDC topics
    • Adams - synchronization essay built around the problems that require synchronization
      • Rosenberg - yup
  • Rao - arch, os, hpc - one of the things that students appreciate is having demonstrations (videos, etc) and practical handouts. Are we going to create some sort of repository of these things.

15:32 : call for proposal (9of12)
short proposal deadline - June 28
response/review to proposal - July 15
submit sections/essays/etc - Aug 22

Objective - get a good chunk done by Fall 2013

• Lee - How do we submit the proposal
  • Prasad - I will set up an easychair
• Yuan - Do we have any template/sample for proposal?
  • Prasad - Not really, this is the first time. It should just be a brief sketch of what want to write.
• Rosenberg - clearly list goals of propsal
• Bogaerts - if we have multiple ideas we should make a separate proposal for each? for related ideas only one proposal?
  • Prasad - more than one proposal from the same team makes sense.
    • Bogaerts - should things that can be adapted to part1/part2 should that be two proposals?
      • Prasad - if it needs to be two then split it, otherwise one is fine.
        • Rosenberg - this is a very informal process
        • unknown - there should be some distinction with what is different between part1 & part2
• Lu - how much should we propose? some things are very broad
  • Prasad - you can propose as much as you want.
• Kothapalli - is there a page limit?
  • Prasad - no
• Brown - is an informal but complete proposal acceptable?
  • Many...yes.
• Adams? - Clearly state the goals but its also not a formal 'bullet-list'. we would like to see it become more collaborative
• Rosenberg - adams to take a look at synchronization essay and become a co-author please?
  • Adams - we can talk about that offline.
• Prasad - I can help with forming teams and facilitating collaboration
• Bogaerts - where can I get the slides?
  • Prasad - I will mail them out.

15:49 CEDR course website
in testing phase now.
can upload stuff. the idea is to make sure that what is uploaded can be accessed/used easily. (so describe correctly)
     Fall-13 round competition deadline June 30, 2013 (going to once a year)

15:51 floor is opened for commends, what worked? what didn't work? workshop?

16:03 - end of meeting.
     Thanks for attending. Great deal of interest

• Adams - likes the idea of a 2-day workshop
  • Prasad - 2-day may or may not be possible. found out that it was very compressed (schedule was tight) can we possibly start sun instead of monday?
    • Adams - that may be better IPDPS would interfere with mon/tues so we HAVE to go to sun/mon with 2-day workshop.
• Marche - webinar problems.
  • Prasad - various wireless network problems, may be solved by going to wired network next time
• Brown - followup on webinar - its appreciated. format with lightning talks - each person giving a poster - this was useful for catching up but made the posters less relevant than in previous years because we got to speak (less visitors to poster). difficult to draw attention to the posters
  • Rosenberg - the conjunction of lightning talks with posters didnt work well? which was better?
    • Brown - the talks were informative, but added time. both don't seem necessary. maybe allow application for one or the other.
      • Prasad - program became tight because of lightning talks, but they seemed beneficial