OGSA Teleconference April 12, 2004
==================================

* Participants

  Dave Berry (NeSC)
  Andrew Grimshaw (UVa)
  Ian Foster (ANL)
  Bill Horn (IBM)
  Hiro Kishimoto (Fujitsu)
  Fred Maciel (Hitachi)
  Andreas Savva (Fujitsu)
  Frank Siebenlist (ANL)
  Chris Smith (Platform)
  Dave Snelling (Fujitsu)
  Latha Srinivasan (HP)
  Jem Treadwell (HP)
  Jay Unger (IBM)

  Minutes: Andreas Savva

* Early discussion

** Teleconference minutes April 7 - Approved 
   
   (With a correction to a participant's name)

** Agenda bashing

  - Re-schedule Information services discussion since Abdeslem Djaoui is not
    present. 
  - Re-schedule PE for next Monday

* Infrastructure discussion

  - Ian sent draft out to list before teleconference.

  - The purpose of this text is to detail assumptions the group is
    making about infrastructure in order to make clear what is
    chosen. Text will be worked into the OGSA document.

  - Rejected proposal to include WS-Addressing in this document
    - OGSA depends indirectly to WS-Addressing via WSRF
    - WS-Addressing has not been submitted to any standards body yet.
   
  - Make clear that everything listed is not pervasive; might have
    cases where something is not being used. (XML example: Might not
    use in places due to performance considerations, e.g., native
    access to data etc.)

  - Include Nataraj Nagaratnam's and Marty Humphrey's comments posted
    to the list? 
    - Marty's text is ok for Frank. 
    - About WS-I security profile, is there a requirement to choose
      only stable ones? No.

  - Frank to send comments on security section to Ian.
  - Jay to send comments on security section to Ian.

* UML

  - Aim: How to express Chapter 6
    - How to represent things, expecially WSRF ResourceProperties.
  - Presented a detailed (slide 2) and a less detailed (slide 3)
    version including a crash course on UML.
  - Clearly using UML for XML representation is not straightforward.

  - Potential problem: audience may not be too familiar with UML
  - Less detailed option is clearly more attractive especially since
    the OGSA document does not have to go to a high level of detail

  - Agreement that we should give it a try; but need consistency in
    the document.

  - Note: The main modelling assumption is the link of state and
    portType/interface
    - See (anotherInterface) and (browseFilter, cursorPosition)

* Glossary

  - At the moment part of the glossary is formatted as a table, part
    as concept pages. Is it easy to maintain?
    - Concept page is good for providing text to other groups, e.g.,
      CDDLM.
    - For the OGSA glossary a table should be enough.

  - Agreed to stay with just the table format and not go into too much
    detail in the definitions
    - The document that defines the term should cover more details anyway.

  - Need to re-check Provisioning concept page and send to CDDLM soon
    because of upcoming deadline on CDDLM foundation document.
    - Structure may be a bit funny: CDDLM foundation document has
      brief outline with reference to OGSA glossary. OGSA glossary
      points to OGSA document.

  - Jem has added Grid and VO definitions to the glossary.
    - Check and comment.

* Provisioning

  - Reviewed briefly last call. This revision does not address all
    comments yet. Some modifications to the first page to address a
    couple of points only.
  - Using PE terminology at the moment.
  
  - First paragraph states "...required for jobs...": It is too
    restrictive since provisioning activity is more general.
    - Agreed to re-phrase along the lines of:
    "...required for executing an activity or function and allocating
     .... in support for that execution."

  - Provisioning is only for active entities? 
    - Argument that passive entities (e.g., storage) might also be
      provisioned ended in agreement that "storage provisioning" means
      provisioning a capability.

  - Agreement that there are many such cycles, not just one giant
    cycle.

  - Different ways of looking at the activities described in the text
    1). Demand of capability
    2). Monitoring an ongoing activity and adjusting (controlling)

    Activity (1) can be rendered as one case of (2).
  
  - So is this really a provisioning loop or is provisioning an
    activity triggered by a control loop like the one described here?
    - IBM MAPE loop example: provisioning as one action triggered by
      the autonomic loop
    - The kind of control loop described here probably forms one
      generic pattern in the architecture. Other examples might be
      fault recovery, etc. 
    - At the moment do not go down the path of trying to render things
      using the 'generic pattern' idea.
    - Back away from the provisioning cycle definition and recast
      provisioning as a straight line activity kicked off possibly
      from a control loop or triggered by some other activity (e.g.,
      job submission).

  - Perhaps recast the cycle as a "service level attainement loop"
    which triggers these activities to maintain the QoS of the
    service.

  - Agreement to rename the section from provisioning to something
    else ("Service level management loop?") and describe provisioning
    as an activity covering only steps 2, 3 of the cycle. Also explain
    that this kind of control loops exist in other places.
    - This provisioning definition is closer to what Jem had in the
      glossary draft sent out for the last call.

  - In addition
    - Identify difference with workload management. Mainly a scheduling
      action.
    - Identify and describe 'disciplines' discretely and make sure
      that it is clear how they are related.

  - Need to sync with provisioning definition in CDDLM; they have
    definition of provisioning along the lines of the cycle presented
    here.

  - Hiro/Andreas prepare next version for Wednesday's call.

* Wrap-up & AOB

  - Next call is for self-management but probably defer since people
    are not ready.