This is a summary of Monday's AG meeting on the Visualization 
Frameworks document.  Thanks to Tom Coffin for taking copious notes.

The participants were;
	John Shalf         LBNL	
	MIke Papka         ANL	
	Tom Coffin         NCSA
	Dave Semararo      NCSA
	Bill Sherman       NCSA	
	Ken Brodlie        U Leeds	
	Stuart Charters    Durham
	Mark Hewitt        Newcastle	
	Ieuan Nicholas     Cardiff
	Nick Avis          Cardiff
	Jim Ahrens         LANL
	Andre Merzky       ZIB	
	Andrei Hutanu      ZIB
	Sarah Bunn         Caltech (phone)	
	John McCorquodale  Caltech		

Intro---------------------------------------------

This meeting was a discussion of visualization, data analysis, model 
structures and the analysis of the additional requirements imposed by 
the movement to a grid-based visualization and data analysis system.  
The GGF wants to see documents that identify best-practices in the 
community to point the way for future grid technology development 
efforts.  Therefore, this effort will revolve around creating such a 
document revolving around data models and requirements for grid vis.

There is an incredibly long history regarding creation of vis model.  
At a minimum, we need a proper assessment assessment of what has been 
done in the past in terms of both theory and implementation.  In the 
DOE, there is an effort to address these issues for interoperable 
visualization and data analysis components called DiVA (Distributed 
Visualization Architecture)
	http://vis.lbl.gov/diva

Likewise the Storage Resource Management (SRM) ISAC in DOE has a 
similar effort afoot for distributed storage management systems.
	http://sdm.lbl.gov/~arie/sdm/SDM.Framework.wshp.htm

This effort will likely focus on the bigger issues that will drive both 
the SRM and DiVA activities.

So, the initial proposal is that the document consist of the following 
4 chapters.
1- data model requirments
2- history of theory in vis modelling
3- history of data model implementation
4- additional reqs. for the grid (distributed parrallel)

The goal is to;
	* understand and document the requirements of the scientific community
	* document the breadth of solutions (both in theory and 
implementation) that attempt to address the needs of the community
	* understand and document the constraints parallel/distributed 
environments place on these data models

All of this is done with the intent of providing more breadth to future 
efforts on Grid-based data analysis environments.  There is a general 
concern that without such a document, the resulting grid services 
standards developed through the GGF will be too narrow in focus to form 
the basis of a robust and broadly-applicable Grid "infrastructure."

----------Discussion------------------

At this point, we turned to each site to get their opinions on what the 
overall content of the document should be and what areas they could 
contribute to such a document.

Ken Brodle:  UK Leeds
	* concerned about the desired result of this document (why are we 
doing this?).
	* also concerned because many people have tried and failed at this 
before.
	* can contribute to the following subjects
	-critique of some past attempts at unified data models
	-requirements that drive development of GVis
	-alterations/improvements/problems with NAG Explorer data structures 
that make them inappropriate for use on the Grid (and some desired 
solutions/requirements that would address these problems).
	
Nick Avis, Leuan Nicholas:  Cardiff (Covise)
	* starting work on distributed environment.  Will share preliminary 
requirements for distributed Covise.  Namely, there will be a 
description of what aspects of the distributed environment will force 
reworking/adjustments in the Covise data structures and abstract data 
model.
	
Stuart Charters at Durham:
	* Interested in contributing a section on issues related to streaming 
data.
	* One issue in distributed environments is how to efficiently manage 
concurrent access to data.
	* partial results out of the pipeline before you have finished.
	* Several peer-to-peer items to the same data source.
	* Scientists and for information visualization for textual information.
	* Will describe both requirements that these use-cases impose on the 
abstract data model as well as implementation issues that address these 
constraints.
	
Jim Ahrens: 	
	* Like many others, feels this will be very *hard* to do.
	* In the interest of focusing the document, there is a general desire 
that we focus on distributed/grid computing issues so that we don't end 
up with an unmanageably large scope.
	*  Interested in considerations of distributed issues for data models 
-- what extensions are required to manage distributed data and how they 
relate to pipeline design.

Andre Merzky and Andrei Hutanu: ZIB
	* They have an interest in the fiber bundle approach as a means to 
provide a more generality as well as adaptors for domain-specific data 
models for efficiency.
	* There is an effort in GGF to tag a file format, but it is in need of 
supplemental information that can be provided through understanding of 
the data models and requirements.
	* Also interested in HDF5 file formats and data models, data sturcture 
requirements for remoteHDF5, fiber bundles.
	* Possible areas of contribution (depending on time)
		-considerations for both fiber bundle data models
		-description of the domain-specific adaptor approach 
(compare/contrast)
		-describe data model requirements of Amira customers such as medical 
imaging and astrophysics research
		-AMR data
		-considerations for distributed pipelines

Dave Semeraro at NCSA:
	* Requirements from SWOF and Teragrid projects
	* Can pull some requirements and ideas from Bob Wilhelmson's ambitious 
with big data movement for weather modeling.  This will provide a very 
good example/use-case.
	* Can also describe requirements for projects in Astrophysics,
		-FLASH code (some AMR)
	* SWOF project wants to deal with data in the Access Grid
		-It would be very interesting to hear what kinds of constraints the 
AG or AG-hosted visualization places on data analysis data models.

Mike Papka:  	
	* This is going to be hard, but it is also necessary.
	* need focus and deadlines.
	
John McCorquodale at CalTech:
	* Research on high throughput parallel data analysis system
		-T221 backed by 4 nodes with a pair of myrinets
		-12-16 GigE channels to Teragrid router
		-1-2Gigabytes/sec from a remote renderer
		-ANL teragrid vis cluster
	* Software support for this environment requires custom streaming 
media protocols (hybrid unicast/multicast)
	* The performance requirements motivate him to create the simplest 
wire format possible for a file  (one serialization format/protocol per 
use).  Generalization is likely to lead to inefficiency.
	* Proposes that the nucleating thought around which we build this 
document is "what requirements are of parallelism" and "what are all of 
the kinds of parallelism that this format percludes".  This will help 
focus the scope of this document.


Action Items:
	Questionnaire:  Tom Coffin suggested we develop a questionnaire to 
determine the disposition of the various people contributing to this 
document and help separate areas of inquiry that participants believe 
are "pressing issues" from more mundane considerations.  We will try to 
get this out to you in a week or two.
	Initial Outline:  A draft outline of the content for this document 
will be created in response to the questionnaire results.
	Milestones:  At that point, we must also need to agree on a schedule 
for completing this document so that it doesn't drag on forever.  This 
is also forthcoming.