Minutes from the UPDT session at GGF9 (Chicago, 10/5-9, 2003)
Tuesday October 6, 10am-11:30am

* Discussed progress on deliverables from our charter
* IP policy reminder
* Review from last meeting at GGF7
* Roadmap for programming on the Grid
     * Define UPDT's role in the larger context of Grid programming
     * GGF8: discussed how we and other people/groups should/could contribut 
    * Need to define the big picture and choose a subpart for UPDT
* GGF8: co-sponsored a workshop with apps-rg group. Very successful

Roadmap:
--------
* Split the big picture into smaller pieces
* Thilo was very interested
* Multiple group/people should be involved in the roadmap
* AMPE should be the owner
* Action item: Bob will send out an email to the chairs in AMPE and the area directors to discuss the roadmap idea.

Status of the User Survey
-------------------------
* Reporting back on Grid results. Only a subset of the slides is presented here.
* All slides will be in Gridforge. 
* Goal: Understand how people develop programs for large-scale systems and the Grid
* Went over the high level description of the survey
* Summary of the results
* Platforms conclusion
* Design cycle
    * Scalability clearly important
    * How to present the graphs of results for Grid
    * Redesign: Low priority. Why?
      * People working from scratch 
      * Low # of people doing Grid work
    * Single login did no get a high priority. Maybe the respondents didn't 
        answer the expected question.
* SDE 
* Debug cycle
    * Print statement -> no steep learning curve
* Tuning cycle
* Maintenance and admin
   * Least filled out 
   * People don't care about this part
* Conclusion

Discussions:
Q: If people fail on local machines what can we do on the Grid?
Other phase: Testing: regression testing, etc. in the Grid env.
What applications were people responding about>
What do production runs look like?

Volunteers to help complete additional surveys and write-up the informational GGF document.

Known problem survey
--------------------
* Went though the survey
* Do you agree with the goals?
Q: Where would application framework go?
   Will repackage the survey a little 

Questions we should consider adding to the survey
Q: Add project/team size (scope of the effort)
   Indentify team members

Q: Commercial versus scientific 
   Quality of the code
   Building a real product or a prototype

Q: Are people managing their own Grid or are they users using somebody else' Grid?

Q: What is the end user's goal?

Discussions:
------------

Potential workshop with apps-rg for GGF10
A lot of interest in a workshop at GGF10
Need volunteers: sign-up sheet



GGF11. A workshop would colide with the CLADE workshop. (Challenges of Large 
applications on distributed environments)

Will send a note to the mailing list about a possible 1/2 day workshop at GGF11

Q: What are the next steps after the surveys for UPDT?
Charter: 2 informational documents and workshop

Wrap-up
-------

User Program Development Tools (10am - 11:30am)

  Review of UPDT at GGF7 (last RG meeting) 
    steering committee wanted some clarification on goals of group 
    ggf7 how app programing would be done and how the group could contribute to 
    these goals 
    "here is the big picture" and in a 2 year time frame here is the sub-part 
    that we will focus on 
    after GGF7 the group was approved 
    GGF8 co-sponsored a workshop with Apps group on "Grid applications and 
    programming tools" 
    this was very helpful 
  "Roadmap" for grid programming 
    roadmap is owned by APME and not UPDT 
    brainstorming was done at ggf7 
    Is anyone interested in this? 
    Comment from Theo: People in data area are doing a similar exercise and he 
    thinks this would be a good idea

    There was some discussion about who groups would be covered by this road map

    There is no area meeting this GGF, which would be a good place to bring this 
    up

    Susanne and Rob will take this up with area directors 
  Preliminary results GGF UPDT user tool survey 
    comment from Ed, not enough grid users, so we should include large scale 
    systems.

    More results than can be covered in 25 minutes

    Results will be posted on GridForge 
    Goal: understand how users develop programs for large scale systems, for the 
    grid, other 
    Survey was distributed to a wide selection of mailing lists, PSC, 
    colleagues, etc. 
    Surveys can still be submitted 

    There were 8 key sections to the survey 
      User profile 
      Platform targeted 
      Design cycle

      Program development cycle

      Debugging cycle 
      Performance Tuning cycle 
      Maintenance and administration 
      Other 
    2/2003 - 9/2003 - 46 respondents 
      20 completed the GGF survey 
      26 completed a similar survey by Dr. Song from NTU 
      ***The data only take the 26 NTU respondents into consideration if their 
      responses do not bias the results.   This is noted in the results (/20 
      indicates GGF survey, /46 indicates both)

      Results can be viewed online, so only the highlights pertaining to grid 
      programming were covered.  There are many more interesting results in the 
      data and Susanne encourages everyone go to look at the data on GridForge 
      and generate some discussion on the mailing list 
      There was some possible misunderstanding regarding the term "Heterogeneous 
      systems".  Intent was that heterogeneous means "application runs on 
      multiple platforms at the same time" 
      Some confusion about definition of "Design cycle".  Intent was the process 
      of conceptual design, not implementation

      Tools used in design cycle: matlab, uml and html

      Susanne asked the group to provide feedback on how can this data be 
      presented better/clearer ? 
      Heterogeneity was the primary priority in the design cycle 
      Question - why was grid redesign rated such a low priority 
        Many people just re-compiling with Mpich-g2 
        Many people likely developing from scratch 
        Could "heterogeneity" be misunderstood (example discussed was single 
        sign-on) 
        Improvement and comments will be incorporated into revised survey

      Software Development - respondents are using the common programing 
      paradigms (see Grid Primer, Lee et. al.) 
      Some discussion on numbers - 35% are grid enabling, of these 20% using 
      Mpichg2 and 25% using Globus 
      Debugging - debug the same way as we debug multi-process programs.   
      Comment from audience...there is no step learning curve to print 
      statements 

      Tuning cycle - everyone has their own technique. 

      Maintenance and administration - lowest response rate. 

      comments on survey 
        There were many things that were perceived, but now we have data to 
        prove these viewpoints 
        People are still thinking about moving applications to...