OGSA session 5
==============

30 January 2007

Introduction
============
- Andrew Grimshaw introduces the audience to the topic and session gals
- Andrew presents his proposal on tackling workflow:
  WFDL (WorkFlow Description Language)

Open Discussion
===============
- At the very least inter-job dependdencies are necessary
  -- Usually Job A creates something that Job B needs so that Job B must run 
     after Job A hass finished
  -- Job A may assert a resource into the system that models its outcome
  -- Job B lists that very resource in its requirements that must be satisfied
     for Job B to run
- Workflows are often similar, or can be seen as programming models for the Grid
  -- This would be out of scope for OGSA
- The reference model may be a place to look at for help and information
- Real Grid applications need to be addressed, noot just a Grid-ified 
 'hello world'
- Co-scheduling is certainly out of scope (as this would be a programming model)
- Requirements and modelling sessions may be required
- Unicore features pretty much the same unwritten requirements set, just encode 
  in Java
- Is workflow in its first simplisting incarnation a language/mechanism to 
  express workflow comsumed by human beings?
- How to layer existing workflows onto OGSA
- 3 initial big choices:
  a) should we stay away?
  -- If not, should we 
     b) collaborate with existing workflow standards bodies,
     c) spin off alone
- straw man draw on alternative a)
  pro: 0
  con: uncounted majority of the audience
- Three posible tracks to follow from now on:
  1) How to layer existing WFs ontop of existing OGF specs
  2) Change JSDL, BES, RSS etc to spport workflow
  3) Interop/collaborate with existing workflow standards bodies
- Diffferentiate between 
  -- workflow description language, and
  -- workflow enactment framework
- Somebody should extend OGSA EMS arch to fulfil the "60% coverage" case
- Volunteers for the three tracks (see above):
  -- Track 1:
     --- OMII UK
     --- IBM              (Fabio Benedetti)
     --- Imperial College (Steve McGough) (lead)
     --- Cardiff University
  -- Track 2:
     --- UVa     (Mark Morgan)
     --- FLE     (Michel Drescher) (lead)
     --- UoM     (Donal Fellows)
  -- Track 3:
     --- delay until 1 and 2 are done
- Suggestion on number 2
  -- have phone confeerences and presentations
  -- scope guidance needed
  -- use cases and workflows needed
- Suggestion on number 1
  -- gridworkflow.org
  
Scoping track 2
===============
- Notion of producing a resource at a location as outcome 
- Explicit data movement jobs in the "input workflow" is out of scope

Any other business
==================
- The SAGA-RG and SAGA core WG would like to be kept informed on the 
  progress about Workflow