August 7, 2014

DOE Highlights Exascale-Focused Research

Tiffany Trader
ASCR logo

Every year the Department of Energy Early Career Award provides outstanding scientists who are early in their careers with funding of at least $150,000 per annum over a five-year span. The 2014 funding round, the program’s fifth, awarded research grants to 35 scientists – including 17 from DOE’s national laboratories and 18 from US universities. The program is designed to bolster the nation’s scientific workforce by supporting exceptional researchers during the crucial early career years, when many scientists do their most formative work.

One of the six primary program tracks is Advanced Scientific Computing Research (ASCR). [The others include Biological and Environmental Research (BER); Basic Energy Sciences (BES), Fusion Energy Sciences (FES); High Energy Physics (HEP), and Nuclear Physics (NP).]

An article at the DOE Office of Science website showcases the important research being conducted by three 2014 ASCR award recipients, whose projects include the following:

  • Statistical Methods for Exascale Performance Modeling
  • Scalable and Energy‐Efficient Methods for Interactive Exploration of Scientific Data
  • Advanced Methods for Immersed Domain Multi‐Physics Computations

With exascale computing poised to appear sometime near the 2020 horizon, these early-career researchers are concerned with overcoming the difficult obstacles that impede this advance. One area of focus is software and the need for elaborate programs that can harness on the order of a billion cores while navigating complex memory systems, networks and accelerator technologies.

Computer scientist Todd Gamblin, the driver for the first project, is working to adapt simulation codes to run efficiently on future exascale machines. Based at the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory in California, Gamblin is developing predictive mathematical models that make this tuning process somewhat less tedious. His research, he says, “explores ways to build machine-learning techniques that predict the causes of performance problems and how to improve them.”

He adds: “We’ll start by building a prototype that predicts what we know already, and we’ll apply that to discover new performance effects in uncharted territory.”

The second project on the list is led by Dr. Florin Rusu, assistant professor with the School of Engineering at the University of California, Merced. Recognizing the ever-worsening problem of data bottlenecks, Rusu is investigating innovative methods and algorithms for interactive exploration of scientific data. His focus is on the interactive exploration of data in order to minimize data movement.

Says Rusu: “You may have lots of data being analyzed in a program that runs for a long time, and you’re not even sure what you’re looking for. In interactive exploration of the data, we design methods that allow us to verify hypotheses much faster.”

The final project concerns the development of exascale-ready computational algorithms, specifically multiphysics problems, which incorporate multiple physical processes. Project lead Dr. Guglielmo Scovazzi, associate professor in the department of civil and environmental engineering at Duke University, states: “Some of the challenges can be associated with the geometrical complexity that arises in high-resolution computations. In fact, building the computational grids required in simulations can create very complex scenarios, especially when you want to work at scale.”

The area of multiphysics is especially relevant to the Department of Energy, since fluid/structure interaction is a prominent feature in wind energy and nuclear reactor systems.

For fluid/structure interaction problems, it is common for the various elements to have different modeling requirements and time scales, which makes standard mesh generation techniques impractical. The research abstract for this program states that “these complex fluid/structure interaction problems will be attacked by means of new immersed boundary and embedded discontinuity methods, in which the fluid and solid domains are discretized using non‐matching grids.”

Exascale computing will be a huge boon to humanity, enabling unprecedented modeling and simulation capabilities, however there is still a lot of work to be done to ensure that applications can take full advantage of these expensive machines.

Share This