Building a better paintbrush
for visualizing a universe of data
Posted June 28, 2011
A picture may be worth a thousand words, but when a star explodes,
a flame burns or cells divide uncontrollably into a metastatic cancer,
a picture also is worth millions of numbers – dimensions,
time, tension, temperature, friction and myriad other variables that help
form vivid images and tell a complex story.
More formally known as visualizations, these representations “have become an indispensable tool,” says Kenneth Moreland, a Sandia National Laboratories (SNL) computer science researcher who builds the latest software-driven methods to render visualizations from mountains of raw data. “Visualization broadens human understanding.”
Moreland (see sidebar, “New ways of seeing at home”) is a principal investigator at the Institute for Ultra-Scale Visualization, or UltraVis, a branch of the Department of Energy’s Scientific Discovery through Advanced Computing (SciDAC) program.
“Ken has made particular contributions to the advancement of parallel visualization,” says Kwan-Liu Ma, UltraVis director and University of California, Davis, computer science department chairman. “He has also been disseminating and deploying our research innovations.”
Moreland is helping tackle what Ma calls “the most challenging problems facing scientists who use the most powerful supercomputers in the world to study the most difficult and important problems in science” – by developing new visualization techniques in climatology, astrophysics, combustion research, fusion, material science and other areas. “As almost all fields of study become more data driven, scientists should use visualization to validate results, discover previous unknowns and communicate their work.”
Besides UC-Davis and Sandia, the UltraVis partnership includes Argonne National Laboratory (ANL), the University of Tennessee and Rutgers and Ohio State universities.
Moreland says he and his coworkers have spent the past decade building parallel computing algorithms that still are evolving. “Our current tools demonstrate excellent scalability on today’s high-performance platforms,” meaning they continue to operate efficiently even as the computers they run on get bigger. “But the nature of parallel computing is always changing.”
