CS-SWIM Meeting Planning
6 March 2007
Short term pre-PSACI
Users/advisory panel identified, invited (CS
victims needed; Al Sussman, Andrew Seigel)
Schedule for PS branch.
Request for capabilities, by 18 March
Branch 25 March or later
Need to have unbroken codes/executables for
PSACI runs
Longer term schedule of PS upgrades (and
concommittent requests for changes)
Short term pre-PSACI
Primary science-driven sims needed
ITER start-up (run on PPPL mhd)
Porcelli test: Linear MHD (run on PPPL mhd)
Rate of tail formation on CMOD (requires Jaguar)
At least first two require EPA component, need
to get this working with current driver/fmwk
Propogation of time-step backtracking not essential
(but we still must allow for it in fmwk and driver)
Short term pre-PSACI
Have portal that
Provides active monitoring of job events
Monitoring via ElViz of desired physics thingies
Allows access to data management system for
users to filter (via easily generated provenance
data) among files generated during a SWIM run
Job launch on PPPL systems from portal
Longer term to address others, esp Jaguar
Short term pre-PSACI
Getting SWIM framework, driver, and
configuration system documented, design
explained
Need schedule for when runs must be
completed for analysis, and who does the runs
(DB will propose this schedule)
Mid Term: between PSACI's
Need capability for signaling between
components (including driver component).
Need use scenarios delineated!
Will only need (in fast MHD) to signal from a component
when it completes, or from a component in mid-
execution?
Will a component need to signal in mid-execution for a
service that it needs before it can continue?
Can use (untested) exception model from CCA for this
Mid Term: between PSACI's
Refine data management system
Get what data files, what data objects are of interest
For monitoring
For ease of extraction later
More generally, use cases for portal and use of
data by physicists here
Identification of most useful metadata (in end user
terms, what kind of questions do you want to ask to
locate data from SWIM runs?)
Don: find the sim I did for a similar case, find the
experimental data used with/for it.
Mid Term: between PSACI's
Science runs needed for PSACI 2008
Timelines like for this year's to get geared up earlier
Need to fulfill promise to effectively use the
Leadership Class Facility
Address job launch from portal to LCF systems
Is this useful, or stick with having LCF runs done
separately (but sending data/monitoring to portal)
Issues of LCF systems; check particularly the
PS being built since it is central.
Mid Term: between PSACI's
Regression tests for components
Identification of a few test cases that can be used
by CS people. Need not be complete coverage.
Performance testing of components (including
driver component)
Build and configure system
Currently a top-level make system set up by Wael
Review is about 1.x years from now. Need to
also determine goals for it. Depends on cycle,
but may need to detail goals earlier for it.