Soaring to discoveries

released 12.01.08

By Barbara Jewett

Computing is changing how we acquire knowledge and express our creativity. The University of Illinois is uniquely poised to contribute to and benefit from this transformation. The new Institute for Advanced Computing Applications and Technologies lets imaginations soar and opens the path to exciting discoveries.

Melting polar icecaps and wild variations in seasonal temperatures frequently are the focus of those discussing climate change. An area equally as affected by climate and other environmental changes, but often overlooked, is our everyday landscapes.

Changing landscapes, and the resulting consequences, are rapidly coming to the forefront as major issues for society. The trajectory of these changes is toward intensively managed environmental systems consisting of built environments—with increasing urbanization, and managed environments, such as agricultural land and artificial wetlands. The scale of these changes has reached the point where they induce significant modification of the water cycle and, as a result, all of the systems that are linked to that cycle such as climate, biogeochemistry, and ecology.

Praveen Kumar and Barbara Minsker, professors of civil and environmental engineering at the University of Illinois at Urbana‑Champaign, and Don Wuebbles, a professor in the university's atmospheric science department, are creating a virtual observatory for sustainability of intensely managed environmental systems (IMES) through the university's new Institute for Advanced Computing Applications and Technologies (IACAT).

"In order to understand the cumulative consequences of changes to an environmental system," Minsker says, "it is essential to examine the interaction between system components, not each component in isolation."

This means their approach needs to go beyond the traditional paradigm and instead provide an end-to-end infrastructure that includes environmental sensing, modeling and prediction, and adaptive management decisions and practices. It will also require supercomputers to process and simulate the data.

Broadening the reach of computing

Going beyond the traditional while expanding upon the University's 50 years of computer science and engineering leadership is the basis for IACAT. By focusing on problems too complex and multi-faceted for individuals or even small groups to tackle, IACAT teams make the most of emerging approaches to simulation- and analysis-based research.

But it's not just computer science and engineering, notes IACAT Director Thom Dunning. "Computing is becoming an integral part of modern research, in all subjects," he explains. "Science, engineering, health care, social science, business, the arts, and the humanities are all utilizing high-performance computing to power breakthroughs in their fields."

The institute is organized into research themes bringing together researchers from a broad array of disciplines with the expertise and computational power of NCSA. Currently there are three IACAT themes, says Dunning, who is also director of NCSA. Each theme includes one or more research projects. They are the Center for Extreme-scale Computing, Advanced Information Systems, and Computing and Creativity. Each of the projects involves six or more Illinois faculty, a similar number of NCSA staff, and several graduate students and postdoctoral fellows. Dunning says the knowledge researchers gain from their projects will be transferred to their respective communities as well as higher education in general, leading to even more discoveries. This will, he believes, have far‑reaching outcomes as we progress through this century, affecting many aspects of our lives.

Powering the arts

Such is the case with the institute's cultural informatics project, which leverages NCSA's high-bandwidth networks and expertise in collaborations into a cultural collaboratory they've named ArtsGrid. Part of the infrastructure for the ArtsGrid will be a suite of tools for sharing aural, visual, and kinetic experiences over networks.

Co-investigators are Michael Ross, director of the university's Krannert Center for the Performing Arts, art and design professor Donna Cox, who also leads NCSA's Advanced Visualization Laboratory, and Illinois' music professor Guy Garnett. In addition to ArtsGrid, they hope to create a high-end laboratory for exploring, integrating, controlling, and developing new technology for performance. Performing arts have embraced technological advances in recent years, including new ways to generate and process audio and video, both offline and during live performances, but most of these new capabilities are hard to use, not widely available, and not usually reliable and durable for production and touring.

The team has an ambitious list of additional topics they want to explore as IACAT researchers. All lead to enriching culture through the production of innovative, leading-edge art that will be recognized worldwide.

"We know that we need to get as much art onto this planet as quickly as possible," says Ross. "Big art, small art, art on our stages and in our museums, art on laptops, in the airwaves, and on the street. We need art that is made with traditional mediums, but we also need art that is yet to be imagined—and can only be imagined in collaboration with unsurpassed technological capacity and the creative minds behind it."

To petascale and beyond

Researchers involved in the three projects in the Center for Extreme-scale Computing theme are focusing on the development of applications and technologies to realize the full potential of petascale computing—and beyond. Their efforts will be used in many cases by Blue Waters, a sustained-petascale computer for open scientific research that is expected to be online in 2011. The machine, funded by the National Science Foundation, will be located on the Illinois campus and is being built by the University of Illinois, its NCSA, IBM, and their partners in the Great Lakes Consortium for Petascale Computation.

Empowering science and engineering researchers by enabling their applications to run 100 times faster—and at much lower cost than on other processors—is the goal of professor Wen-mei Hwu of the university's electrical and computer engineering department. Through the institute, he'll expand his work at the university's Coordinated Science Laboratory as he and his collaborators develop science and engineering application algorithms, programming tools, and software, for deployment of next-generation many-core processors like graphics processing units (GPUs). Hwu says GPU computing will lower hardware costs, while the speed and power will allow explorations considered unfeasible today. His team is tailoring experimental systems and developing software to best enable science and engineering breakthroughs.

Hwu says his team is seeing some GPUs with very impressive "speeds and feeds." However, as with any parallel computing hardware, the most important measure is the application's performance and the programming efforts involved to achieve that performance level.

Easing the programming effort required to achieve high performance levels is the goal of IACAT researchers Laxmikant (Sanjay) Kale and Duane Johnson. With new applications using sophisticated, multiscale, multiphysics, dynamically adaptive algorithms, Kale notes it's challenging to harness the in-service compute power of applications efficiently.

The need for computer scientists and domain scientists to work together is why Kale, an Illinois computer science professor, and Johnson, a professor in the materials science and engineering department, joined forces. Together they'll focus on developing petascale applications in the fields of astrophysics, biomolecular science, and materials science, while conducting complementary computer science research on the tools and technologies needed to support the development of these applications.

Applications that run on petascale machines are only part of the knowledge discovery story. Many applications in science and engineering require multiscale simulations to accurately describe all of the key phenomena. According to IACAT researcher Richard Braatz, a professor of chemical and biomolecular engineering at Illinois, the number of research problems being tackled by using multiscale simulation is growing rapidly. This growth, he says, is consistent with a 2006 report from a National Science Foundation panel that identified multiscale simulation as the number one challenge in advancing knowledge and understanding in simulation-based engineering science.

"Multiscale simulation can be defined as simulation over many orders of magnitude in time and length scales," he explains. "Different simulation methods are most effective at different scales, and the coupling of these multiple methods complicates the design of algorithms for numerical computing."

Braatz and his team hope to create a universal set of algorithms, software, and data analysis tools for multiscale simulation with validation in applications for which increased predictability matters. This will make multiscale simulation accessible to, and feasible for, a wide variety of researchers across a broad spectrum of science and engineering communities.

More information on IACAT and the research projects can be found at: www.iacat.uiuc.edu.