Intelligent Light and FieldView

In Situ Data Management at DOE Workshop and SIAM CSE

​As CFD datasets get larger and unsteady computation becomes more common, file I/O is an increasing bottleneck. Methods that write post-processing extracts directly from solvers as they are running have proven to be a solution to this problem. Intelligent Light has long advocated for this approach and recently we were asked to participate in activities to advance the state of the art in this area.

In January, I participated in the DOE's "In Situ Data Management (ISDM)" workshop which focused on defining areas for research intended to ease the analysis and understanding of massive simulations enabled by exa-scale computing. A workshop report is forthcoming.

In February, I was invited to present at a mini-symposium at the SIAM CSE 2019 meeting aimed at avoiding the big data problem through in situ techniques.

If you would like to know more about In Situ Data Management and CFD Data Analytics or to schedule a meeting with Steve Legensky, send email to datascience@ilight.com.

Data from Juan D. Colmenares, Svetlana Poroseva, Yulia T. Peet, and Scott M. Murman. "Analysis of uncertainty sources in DNS of a turbulent mixing layer using Nek5000", 2018 Fluid Dynamics Conference, AIAA AVIATION Forum, (AIAA 2018-3226). Simulations were performed on the Pleiades Computer system at NASA Ames Research Center. Images created by Intelligent Light.


Continue reading
  0 Comments
0 Comments

HPC, In Situ and Data Analytics at SC18

Advances in the use of HPC, In Situ and Data Analytics for CFD dominate our contribution to SC18. Join us at events, in workshops and technical sessions and on the exhibit floor in Booth #826 to learn how these advances will help you.

Data Science Meets CFD - Steve Legensky, Invited speaker at In Situ Analysis and Visualization Workshop

VisIt Users and Developers Reception - Get an update on VisIt 3.0 and VisIt Prime

SENSEI Tutorial: Cross-Platform View of In Situ Analytics

Data from Juan D. Colmenares, Svetlana Poroseva, Yulia T. Peet, and Scott M. Murman. "Analysis of uncertainty sources in DNS of a turbulent mixing layer using Nek5000", 2018 Fluid Dynamics Conference, AIAA AVIATION Forum, (AIAA 2018-3226). Simulations were performed on the Pleiades Computer system at NASA Ames Research Center. Images created by Intelligent Light.


Invited talk: Data Science Meets CFD, Steve Legensky, Intelligent Light

Presentation to the ISAV 2018 Workshop: In Situ Infrastructures for Enabling Extreme-Scale Analysis and Visualization

Monday, November 12, 3:30-4:25PM, Room D168

Advances in computational fluid dynamics (CFD) and high-performance computing (HPC) have allowed an amazing increase in model fidelity. For CFD practitioners, these new capabilities present challenges with the analysis of highly unsteady flows of ever-increasing complexity. Simultaneous advances in data science offer promising techniques for resolving these issues that are just starting to be applied to CFD post-processing and knowledge extraction workflows.

ISAV 2018: In Situ Infrastructures for Enabling Extreme-Scale Analysis and Visualization

Session: Monday, November 12, 9:00-5:30PM Room D168

The considerable interest in the HPC community regarding in situ analysis and visualization is due to several factors. First is an I/O cost savings, where data is analyzed and visualized while being generated, without first storing to a file system. Second is the potential for increased accuracy, where fine temporal sampling of transient analysis might expose some complex behavior missed in coarse temporal sampling. Third is the ability to use all available resources, CPU's and accelerators, in the computation of analysis products.

The workshop brings together researchers, developers and practitioners from industry, academia and government laboratories developing, applying and deploying in situ methods in extreme-scale, high-performance computing. 

Data from Juan D. Colmenares, Svetlana Poroseva, Yulia T. Peet, and Scott M. Murman. "Analysis of uncertainty sources in DNS of a turbulent mixing layer using Nek5000", 2018 Fluid Dynamics Conference, AIAA AVIATION Forum, (AIAA 2018-3226). Simulations were performed on the Pleiades Computer system at NASA Ames Research Center. Images created by Intelligent Light.

Tutorial: SENSEI Cross-Platform View of In Situ Analytics
Presenters: E. Wes Bethel, David Thompson, Burlen Loring, Silvio Rizzi, Brad Whitlock, Matthew Wolf, Patrick O'Leary

Sunday, November 11, 1:30-5:00PM Room C147

This tutorial covers the design and use of SENSEI, a platform for in situ visualization and analysis. Attendees will learn the basics of in situ analysis and visualization — which eliminates the need for writing large simulation state files or other data that prevents scaling to large machines — while being exposed to advanced analysis such as autocorrelation, interactive monitoring and computational steering.

Image courtesy of Michael Brazell and Prof. Dimitri Mavriplis, University of Wyoming

Spend time with our experts:

Come by Booth #826 or request a meeting, to learn and explore how we can help you advance your CFD.

In-booth presentation: Data Science Meets CFD, Steve Legensky

Tuesday, 4:00PM in Booth #826, followed by reception including beer and hors d'oeuvres

Continue reading
  0 Comments
0 Comments

Come See Us at Overset Grid Symposium 2018!

We are happy to be sponsors of the symposium again this year and look forward to seeing you there.

There are three opportunities to meet us and learn more about our work with FieldView, VVUQ and ​Data Analytics:

  1. Tutorial on Monday: FieldView: Parallel Calculation, Remote Visualization, Automation and Extract Based Workflows (See below for details)
  2. Presentation during symposium: Spectre: A Computational Environment for Managing Total Uncertainty Quantification of OVERFLOW based CFD studies  (See below for details)
  3. Visit our table in the exhibit area
We look forward to seeing you there!
Turbulent Wakes of the Lillgrund Wind Farm, simulated with W2A2KE3D, in situ processing with VisIt / Libsim image rendered in FieldView
Data Courtesy: M. Brazell, A. Kirby, and D. Mavriplis University of Wyoming   Image by: Intelligent Light


Earl Duque, PhD Manager of Applied Research, Intelligent Light

Spectre: A Computational Environment for Managing Total Uncertainty Quantification of OVERFLOW based CFD studies

Dr. Earl P.N. Duque, Manager of Applied Research Group, Intelligent Light

The uncertainty in CFD based simulation results may be quantified through rigorous verification, validation and uncertainty quantification (VVUQ) procedures. Procedures and frameworks such as the ASME V&V 20, and Oberkampf and Roy Uncertainty frame work require extensive simulations, post-processing workflows and management of the simulation results to quantify the contributions of numerical uncertainty, model input uncertainty and model form uncertainty to the total uncertainty for any given study or design. To date, the application of VVUQ procedures were ad hoc implementations. Spectre is a new computational environment designed to enable CFD practitioners to easily quantify the total uncertainty in their computational study and campaigns. This presentation will present how Spectre makes use of "Wizard-based" user interfaces to guide the user through all the steps needed to arrive at the total uncertainty using the OVERFLOW2 code. Work in progress case studies based upon a UQ study of a NACA0012 airfoil at zero lift, a multi-element airfoil and the AIAA High Lift Prediction workshop Common Research Model will be presented.

Stephen Makinen, PhD Customer Application Engineer, Intelligent Light

FieldView: Parallel Calculation, Remote Visualization, Automation and Extract Based Workflows

Dr. Stephen Makinen, Customer Application Engineer, Intelligent Light

Synopsis: Modern high-fidelity physics simulation methods generate massive amounts of data, so a challenging scenario exists for efficiently processing results to create new knowledge. The rate of new data generation has far surpassed data rates for disk read/write operations and transfer from remote computational facilities. FieldView post-processing enables scientists and engineers to efficiently create new knowledge with scalable methods that navigate around these limitations. The FieldView tutorial will cover the following topics.

  • Overview of FieldView including new and upcoming features
  • Parallel calculations for efficient post-processing
  • Remote visualizations to reduce data transfer
  • Extract based workflows to minimize disk read/write operations
  • Data analytics for solution decomposition and insight into the underlying state-variables
  • Automation with Scripting
  • Discussion
Continue reading
  0 Comments
0 Comments

Steve Legensky at 70th HPC User Forum

​On Thursday, September 6, I will present "Data Science meets CFD:  FieldView Analytics in Engineering" at the 70th HPC User Forum in Detroit, MI.

I am pleased to have been asked to update the HPC community on our vision for using data science to complement and enhance CFD in science and engineering. This is a very active area of research today. In March Intelligent Light was honored to participate in the "US-Japan Workshop on Bridging Fluid Mechanics and Data Science" in Tokyo, the first such event in history. We have been meeting with our FieldView customers throughout the world to understand their needs and to see how we can help them solve problems through FieldView Analytics. Data science can help us to better understand the dynamical systems represented by CFD simulations and revealed by experiments, with the goal of getting better designs sooner with less computation.

For more info on the 70th HPC User Forum: www.hpcuserforum.com

The Agenda can be seen here: https://www.hpcuserforum.com/downloads/agenda.pdf

Continue reading
  0 Comments
0 Comments

ScienceNode: Wind Farm CFD by University of Wyoming and Intelligent Light

In Situ processing enables full wind farm simulation by reducing file size from 234TB to under 1TB

Excerpted from an article originally published on ScienceNode. Read the complete article

Researchers at the University of Wyoming (UW) and Intelligent Light are working hard to improve the efficiency of wind energy. One aspect they're currently investigating is the use of computational fluid dynamics (CFD) to simulate, predict, and improve the performance of wind farms.

CFD uses supercomputers to model the behavior of fluids such as air or water over moving surfaces. In this instance, UW professor of mechanical engineering Dimitri Mavriplis and Earl Duque at Intelligent Light turned to CFD to study how wind currents transform after coming into contact with a moving turbine's blades in the context of a complete wind farm.

Using the Cheyenne supercomputer at the National Center for Atmospheric Research (NCAR)-Wyoming Supercomputer Alliance, Duque and UW researchers Mavriplis, Michael Brazell and Andrew Kirby were able to model wind farms in unique ways. 

What really makes their research unique is the use of in-situ processing. Unlike traditional visualizations produced from stored data, in-situ processing means that visualizations and analysis can be created as a simulation takes place, without writing to disk.

In-situ processing eliminates file transfer bottlenecks, allows increased fidelity, and faster turnaround. 

"In-situ has helped reduce the amount of 3D data that we need to store. Just the restart files take 10 TB on a 12-hour run. Automation has gone way up; we have scripts that go straight from simulation to animation. Also, in-situ has helped with debugging—I don't need to pull down 3D data, and I can just use slices, save those to a FieldView XDB file, and then view that on my local machine."

Michael Brazell, University of Wyoming

Excerpts trom an article was originally published on ScienceNode.org. Read the original article.

CFD images courtesy of Prof. Dimitry Mavripilis and Andrew Kirby, University of Wyoming, an Intelligent Light University Partner.
 
Continue reading
  0 Comments
0 Comments