application of salssa framework to the validation of...
TRANSCRIPT
Parallel, remote visualization of large data sets with VisIt
in collaboration with VACET and Prabhat (NERSC)
Wide range of tools
VisIt
Tecplot
General Mesh Viewer (GMV)
Plotting tools (gnuplot etc)
Sharing of visualization scripts
Job Launching Submits remote jobs to Chinook,
Franklin, and workstations; easily
supports new queue systems
Launches multiple jobs concurrently,
to multiple machines
Monitors remote job progress with
the option to terminate
Generates boilerplate text for job
submission scripts, which can be
customized on a per-machine per-task
basis
User control over where to store
output files, such as to an archive (left),
to the Content Management System,
or left on the compute server
Implemented with the Kepler
workflow system
Workflow and Tool Integration
Tasks are the execution of tools or
a sequence of tools. Computational
models are treated like tools
New tools are easily added
through a text file-based registry; a
registration user tool is planned
Tools can be executed locally or
remotely
Visualization tools are registered
just like any other task and can be
desktop tools like TecplotTM or those
requiring a supercomputer to run
A programming model is provided
to plug-in custom wxPython GUIs to
allow complex processing or
visualization
Many subsurface flow and transport problems of importance today
involve coupled non-linear flow, transport, and reaction in media
exhibiting complex heterogeneity. In particular, problems involving
biological mediation of reactions and/or mineral
precipitation/dissolution fall into this class of problems. We are
developing a hybrid multi-scale modeling framework that integrates
models with diverse representations of physics, chemistry and biology
at different scales (sub-pore, pore and continuum). The modeling
framework takes advantage of advanced computational technologies
including parallel code components using the Common Component
Architecture (CCA). A pore scale Smoothed Particle Hydrodynamics
(SPH) model has been built within the CCA framework.
The Support Architecture for Large-scale Subsurface Simulation and
Analysis (SALSSA) software framework addresses the macro
modeling processes such as running simulations and analyzing their
results. SALSSA integrates workflow, data management, and
visualization supporting the development, validation, and use of the
SPH model.
Hybrid Multi-scale Modeling Frameworks
Provenance Tracking and Data Management
Automatically keeps track of user activities including location of data files,
associated metadata, and provenance.
Enables generation of multiple views including provenance graph shown
above and simulation table shown below
Supports automatic metadata extraction to create context aware views
Provides clear record that can be used for verification
Record can be used to repeat activities such as validation testing
Flexible metadata model easily incorporates new metadata
Users can dynamically add/remove columns from the table for any
metadata that has been collected
Automatically transfers all inputs for staging to remote workstations and
then stores outputs back to the Alfresco Content Management System
Based on Sesame Resource Description Framework (RDF) and Alfresco
content management system
http://subsurface.pnl.gov/salssa
Data &
Provenance
Store
Franklin (NERSC)
Chinook (EMSL)
Workstations…
High Performance
Storage System
(HPSS)
Job Scripts Tool Registry Metadata
extractors
MIME type
extractors
Custom Tool
User InterfacesStandard tools
User preferred tools
User developed tools
Default scripts
Power user scriptsInput file parameters
Output file results
Extensibility Mechanisms
User Environment
Visualization
The SALSSA framework is comprised of a user environment that
supports the integration of a wide range of tools, a mechanism for
launching simulations to a variety of machines, a sophisticated data
sub-system that manages provenance and keeps references to
simulation data that may be located on archives or compute systems,
and a mechanism to invoke tools located with the data. Scientists can
extend the system to support new tools by modifying data files and
scripts. For more sophisticated integration, a Python-based
programming model is available.
Density distribution of SPH simulation in
porous media
Visualization of SPH simulation
provided by Kwan Lui Ma
(UltraVis)
Data / Job
Submission
Service
The “Organizer” is the primary user interaction tool. It is structured around the concept of a
“study” where a single study is a collection of setup, simulation execution, and analysis tasks. The
data associated with each task is automatically recorded along with provenance which results in
the graph-based view shown above. New tasks can be initiated based on the type of an existing
task and its data. The figure above shows the graph view of the simulation and analysis task
graph for the validation study. A range of graph viewing and manipulation options are available.
Inputs and outputs for each task can be displayed by selecting the task.
Simulation Execution
SALSSA Architecture
The set of available tools is
organized into categories. A wide
range of tools can be incorporated
including: models, visualization
tools, data transformation
tools/scripts, desktop tools, and
custom tools. Users initiate new
tasks by double-clicking on a task
within the tree. If a process or
data is selected, the task tree will
only show tools that can operate
on the specified data.
When no tools are “known”
to handle a particular data
type, the user is prompted
to select a tool. This allows
easy integration of desktop
tools such as users’
preferred editors, plotting
tools, or simulation codes.
A wizard will prompt for any
inputs that cannot be
determined from the user
selection. The prompting is
determined by input file and
parameters specified in the
tool registration file.
The wizard has been
customized to support
editing of SPH input
files and specification
of a range of
parameters for which
separate simulations
should be generated.
Kepler
Sesame /
Alfresco
Application of SALSSA Framework to the
Validation of Smoothed Particle Hydrodynamics
Simulations of Low Reynolds Number FlowsKaren Schuchardt, Jared Chase, Jeff Daily, Todd Elsethagen, Bruce Palmer, Tim Scheibe
About Pacific Northwest
National Laboratory
The Pacific Northwest National Laboratory,
located in southeastern Washington State, is a
U.S. Department of Energy Office of Science
laboratory that solves complex problems in
energy, national security and the environment,
and advances scientific frontiers in the chemical,
biological, materials, environmental and
computational sciences. The Laboratory
employs 4,000 staff members, has a $760
million annual budget, and has been managed
by Ohio-based Battelle since 1965.
For more information about the science you see
here, please contact:
Karen Schuchardt
Pacific Northwest National Laboratory
P.O. Box 999, K7-90
Richland, WA 99352
(509) 375-6525
Computational Model
Smoothed Particle Hydrodynamics (SPH):
A Lagrangian particle method for solving systems of partial differential equations
Chase JM, KL Schuchardt, G Chin, Jr, JA Daily, and TD Scheibe. 2008. "Iterative Workflows for Numerical Simulations in Subsurface Sciences." In Proceedings of 2008 IEEE Congress on Services Part 1, pp. 461-464. IEEE Computer Society, Los Alamitos, CA.
Gibson TD, KL Schuchardt, and EG Stephan. 2009. "Application of Named Graphs Towards Custom Provenance Views." In 1st Workshop on the Theory and Practice of Provenance (TaPP '09).
Schuchardt KL, GD Black, JM Chase, TO Elsethagen, and L Sun. 2007. "Process Integration, data management, and visualization framework for subsurface sciences." Journal of Physics: Conference Series 78:012064, 1-5.
Scheibe TD, AM Tartakovsky, DM Tartakovsky, GD Redden, P Meakin, BJ Palmer, and KL Schuchardt. 2008. "Hybrid numerical methods for multiscale simulations of subsurface biogeochemical processes." In Proceedings of SciDAC 2008: Journal of Physics, vol. 125, p. 012054. IOP Publishers, Bristol, United Kingdom.
Publications
Collaborators
Fluid flow in porous media at velocities common in natural porous media (i.e., groundwater
flow) occur at low Reynolds numbers and therefore it is important to verify that the SPH
model is producing accurate flow solutions in this regime. The SPH model contains
parameters that do not appear in the corresponding incompressible Navier-Stokes
equations and it is important to identify value ranges for these parameters that lead to good
solutions for the flow profiles in topologically complex media. These parameters include a
speed of sound c, which manifests itself by introducing a finite compressibility into the flow,
and a weighting function diameter h that is roughly analogous to the grid spacing in grid-
based numerical methods. Ideally, solutions should be independent of these parameters in
some limit.
The approach taken is to perform a series of simulations using relatively simple geometries
where the flow solutions can be compared against analytic results or numerical results
obtained using other methods, such as finite elements. The two flow geometries are slit
flow between two parallel walls and flow in a cubic array of periodic spheres. The slit flow
problem leads to a parabolic flow profile for which there is an analytic solution and the
periodic sphere problem can be modeled using standard grid-based numerical methods.
Both of the configurations will be simulated using SPH over a range of parameters and
compared to validated solutions.
Validation of Low Reynolds Number Flows
Preliminary Results
Validation of the SPH model is currently under way. Tens of parameter
configurations have already been executed and analyzed within the
SALSSA workflow environment. Results of a selected SPH validation
simulation run are shown above. Left: comparison of simulated velocity
profile along a cross-section through the slit; Center: Time variation of
simulated mean velocity as steady-state flow is approached; Right: Density
variations within the simulated flow field as a function of time. All variables
are presented in non-dimensional units.
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0 5 10 15 20
x
ve
loc
ity
Simulated
Analytical
Solution
0.00E+00
5.00E-03
1.00E-02
1.50E-02
2.00E-02
2.50E-02
3.00E-02
3.50E-02
4.00E-02
4.50E-02
5.00E-02
0 100 200 300 400 500
Time
Mean
velo
cit
y
27.06
27.065
27.07
27.075
27.08
27.085
27.09
0 100 200 300 400 500
Time
De
ns
ity
Minimum
Maximum
Simulation of non-reactive contaminant transport
using SPH framework. Porous media surface is in
green, isosurfaces of velocity are red and
isosurfaces of concentration are blue. The
visualization is done using the VisIt parallel software
package.
The Organizer’s Simulation View provides a spreadsheet view of associated metadata and
provenance information for selected simulation runs. These views are customizable, allowing
a user to add, remove, or reorder the data columns.
PN
NL-S
A-6
6855