Turbulent Flow Simulation at the Exascale: Opportunities and Challenges Workshop

Sponsored by the U.S. Department of Energy
Office of Advanced Scientific Computing Research
Autograph Mayflower Hotel
Washington, DC
August 4 - 5, 2015


Turbulent Flow Simulation at the Exascale: Opportunities and Challenges Workshop

The Turbulent Flow Simulation workshop will be held at the Autograph Mayflower Hotel, 1127 Connecticut Avenue NW, Washington, DC, telephone 202-347-3000. The workshop will begin on Tuesday, August 4, 2015, at 8:00 am and end on Wednesday, August 5, 2015, by 5:00 pm. A registration fee of $138.00 will be charged to attend this workshop. Registration fee includes continental breakfast, lunch, and AM and PM breaks. Workshop check in and continental breakfast will begin at 7:15 a.m.

Meeting Overview

The objective of this workshop is to define and describe the challenges and opportunities that computing at the exascale will bring to turbulent-flow simulations in applied science and technology. The need for accurate simulation of turbulent flows is evident across the US Department of Energy applied-science and engineering portfolios, including combustion, plasma physics, nuclear-reactor physics, wind energy, and atmospheric science. The workshop will bring together experts in turbulent- flow simulation, computational mathematics, and high-performance computing. Building upon previous ASCR workshops on exascale computing, participants will be expected to help define a research agenda and path forward that enables scientists and engineers to continually leverage, engage, and direct advances in computational systems on the path to exascale computing.

The workshop will be structured around parallel working-group sessions separated by a small number of invited talks. Workshop participants are expected to contribute and discuss ideas that will be incorporated into a workshop report. Questions to be addressed include the following:

1. What is the potential impact of exascale simulations of turbulent flow on our fundamental understanding of turbulence?

Most of the recent progress in fundamental understanding of turbulence has come from simulations that push down to the smallest length scales in the flow. In some cases, molecular-level effects have been incorporated. This is an extremely computationally intensive approach. The potential for discovery science in turbulence using increased computing power should be examined in detail.

2. What are the potential impacts on DOE Applied Technology programs (Wind Energy, Nuclear Energy, Stockpile Stewardship)?

While simulations aimed at discovery science in turbulence are generally done using approaches such as Direct Numeric Simulation, engineering simulations in turbulent flows are performed using Large Eddy Simulations and Reynolds Averaged Navier-Stokes simulations. Therefore, the benefits, and potential impacts, of improved simulation capabilities in these areas should be considered separately.

3. What are the potential impacts of exascale simulations that include improved turbulent flow simulations on problems of scientific interest to the Department of Energy?

The potential impacts of increased simulation capability in turbulent flows for climate, fusion, and other DOE Office of Science problems is likely to differ from those of applied programs, and should be considered separately.

4. What are the approaches and challenges to adapting today's software and algorithms for turbulent-flow simulation at the exascale?

The architecture changes created by exascale computing change the relative cost of operations. Memory and chip-to-chip communication capabilities are limited relative to floating point operations. This will create new challenges for implementing turbulent flow solvers that may be particularly relevant in a multi-scale problem such as turbulent flow.

5. What are the opportunities for new software and algorithms for turbulent-flow simulation that may be enabled by exascale capabilities and architectures?

In some areas, the "flops are free" paradigm created by exascale may lead to new computational approaches, and new capabilities. The correct solution approach will not always be a modified version of petascale codes. We propose examining this question for turbulent fluid mechanics.

Program Committee:
Stanislav Boldyrev (U. of Wisconsin)
Paul Fischer (ANL and U. of Illinois)
Ray Grout (NREL)
William Gustafson (PNNL)
Robert Moser (U. of Texas)
Michael Sprague (Chair, NREL)

DOE Point of Contact
Michael Martin (ASCR)