 |
Workers Section |
Issues Raised at the 2002 Sherwood Meeting
The following are Remote/Distributed Computing issues that have been raised
while at the Sherwood meeting
- The intermitent failure never occured. We wonder if the trouble
at TTF was due to a bad PPP port at GA.
- The demo software ran very smoothly, no errors or problems
- Only difficulty was the down time of ESNET on Sunday
- TRANSP runs were quicker
- 4 hours at TTF to run TRANSP, only 1/2 hour at Sherwood
- Did someone speed up the speed of the MDSplus writes? Tina?
- Monitoring
- Used Grid-Info-Search to display data
- It worked well
- Netlogger
- LogDaemon large number of used ports issue resolved
The following are Visualization Computing issues that have been raised
while at the Sherwood meeting
- Better scaling resolution of vector data
- Color map viewing died and SciRUN needed to be restarted
- Better control on color map for alpha values
The following are issues that have been raised while discussing
the Security/Remote Computing demonstration with the scientific audience
- The Sherwood group is composed of scientists who run very large codes
and their user base is small. Thus the Grid concept of computing was
of less interest
- Jeff Candy and the Gryo code expressed some interest to test a simple implementation
- If were were to try to do a bigger code would need to bring
NERSC into the GRID since these codes run on NERSC systems
- Dalton and Nimrod expressed that other things like vis are a higher priority
- Big issue regarding MDSplus storage of NIMROD
- They may start to swamp our storage
- We need to pursue the NERSC red carpet treatment to get them
to run an MDSplus server tied to their HPSS system.
The following are issues that have been raised while discussing
the Visualization demonstration with the scientific audience
- Volume vis is nice for an overall (rough) viewing but they
desire more quantitative information
- 2D slice, high resolution, will give them quantitative information
- Surface plot - Slice the 3D image, e.g. plot (x,y) and make Z the pressure
- Nested isosurfaces with translucency (e.g. SC01 EFIT Vis)
- Better seeding methods for streamlines
- Linear interpolation in streamlines not sufficient
- Have the ability to use the same interpolation as was
done in the code (e.g. MDSplus storage)
- Pieces of software in SciRUN can be compiled on the fly
- Ability to use different integration methods
- Animator to view different time slices at a time for a pseudo-movie
- Can we bring in Microturbulence data (e.g. Candy - Gryo) to be
visualized in the same viewer as MacroStability data
- Benchmark data for validating visualization tool
- Data storage and source issues
- Speed of data retrieval a key
- Get MIT to use the vis tool
- Get MIT to fetch NIMROD data using code (getnim) that is used in SciRUN
- Distribute this Vis tool via an RPM
- Can we make an interface for the expert and the novice - This and the issues
below are general philosophical design concerns that we must address
before continuing.
- Scientists did mention a concern with the learning curve. Network
programing is very new to them
- The novice interface would be point and click. A reduced functionality
but ease of use
- Expert used the network - gets the full power
- Give a general MDSplus interface then the viewer can be used for all data.
Here users are willing to type since they speak the language of data.
