DOE SciDAC Researchers Meet in Virginia for Program LaunchJune 3, 2002
David Keyes, lead researcher of the SciDAC project in applied mathematics known as TOPS---the Terascale Optimal PDE Simulations Center.
Approximately 140 researchers met in Reston, Virginia, in mid-January to mark the start of the Department of Energy's SciDAC (Scientific Discovery through Advanced Computing) initiative. A major investment---the first round of awards, in FY 2001, totaled $57 million---SciDAC reflects DOE's recognition that the extremely complex and accurate models of physical, chemical, and biological systems made possible by terascale computers can be developed only by interdisciplinary teams.
Participants in the 51 SciDAC projects funded to date include application scientists, applied mathematicians, and computer scientists. Working in teams, they will develop the scientific computing software and hardware infrastructure needed to use terascale computers to advance fundamental research in several areas related to DOE's missions. Among the areas targeted by SciDAC are climate modeling, fusion energy sciences, chemical sciences, nuclear astrophysics, high-energy physics, and high-performance computing.
Many of the SciDAC researchers will be developing a new generation of scientific simulation codes that take full advantage of the computing capabilities of terascale computers to address ever larger, more complex problems. Others will focus on improved mathematical and computing systems software that will make it possible to run the new codes effectively on modern parallel computers.
The January meeting brought together principal investigators from each of the 51 SciDAC projects. James Decker, director of DOE's Office of Science, and SciDAC program director David Bader opened the meeting. Together, they provided the context for the SciDAC program by highlighting its relation to the DOE mission and existing research portfolio. In addition, they discussed the unprecedented opportunities for cross-disciplinary research and the challenges facing SciDAC researchers---including the difficulty of forming ongoing collaborations among different communities and the need to instill a culture of accountability without stifling creative research.
"These projects represent a significant change in the way we do computational research, with greater emphasis on integrated teams," Decker says of the SciDAC program. "Our strategy is to support coordinated efforts by the scientists working to solve complex problems in physics, chemistry, and biology, and the applied mathematicians and computer scientists working to develop the computational tools required for that research."
The remainder of the first morning was devoted to presentations by researchers from SciDAC application centers, which cover four broad areas: biological and environmental sciences, particularly climate prediction; high-energy and nuclear physics; basic energy sciences, particularly combustion and chemically reacting flows; and fusion energy sciences. To facilitate discussion among the centers, the speakers from the application areas were asked to delineate the results they hoped to obtain from the SciDAC project's Integrated Software Infrastructure Centers (ISICs, three in applied mathematics and four in computer science).
Among the areas mentioned were mesh generation on complex geometries, new discretization schemes, adaptive mesh refinement, improved linear systems and eigenvalue solvers, scalable algorithms and software, improved visualization, scalable data management, and collaborative technologies. To showcase ongoing SciDAC research funded to meet these needs, the afternoon session was devoted to presentations by representatives of the ISICs, and of the SciDAC centers created to develop software in support of research collaboration; the speakers in this session emphasized tools and technologies that would be of interest to the application teams.
Three speakers described work under way in the applied mathematics centers. The first, David Keyes of the Terascale Optimal PDE Simulations (TOPS) center, discussed the project's efforts to develop, implement, and support optimal or near-optimal schemes for partial differential equation-based simulations and closely related tasks, including optimization of PDE-constrained systems, sensitivity analysis, eigenanalysis, and adaptive time integration, as well as core implicit linear and nonlinear solvers. James Glimm described progress at another applied mathematics center---the Terascale Simulation Tools and Technologies (TSTT) center. TSTT's mission is the creation of interoperable and interchangeable meshing and discretization technologies.
Philip Colella represented the third applied mathematics center, the Application Framework for Partial Differential Equations Center (APDEC). The goal of this center is to insert advanced adaptive mesh and discretization technologies into the magnetic fusion, accelerator design, and combustion application areas. Algorithm development in this center focuses on block structured adaptive mesh refinement with particles and Cartesian mesh embedded boundary methods.
The final presentations of the first day focused on the "collaboratory" software to be developed under the SciDAC program. Using the products of these efforts, geographically separated scientists will be able to use scientific instruments and computers remotely and to work together with distant colleagues as teams, sharing data more readily. The goals of these centers include high-performance middleware services, such as the development of security architectures, storage resource management, scientific annotation, and collaborative environments. In addition, the continued development of the DOE Science Grid will lay the groundwork to support distributed collaborations, very large data volumes, unique instruments, and the incorporation of supercomputers into these environments. Of the several application groups in a position to benefit from usable Grid technologies and environments, the Particle Physics data grid collaboratory project was identified as a particular example.
The second day began with a lively four-hour poster session that encouraged new interactions among the centers. Follow-up presentations, in which the field representatives for applications and ISICs presented new contacts, revealed that each center had found five to ten potential new collaborations for further exploration in the coming months.
Readers can learn more about SciDAC by visiting the Web site, which provides useful information about the program and pointers to many of the individual centers. In addition, the program for SIAM's 50th Anniversary Meeting includes several minisymposia that feature various aspects of the SciDAC ISICs in applied mathematics. Charles Romine, DOE program manager for the applied mathematics ISICs, for example, organized a session that will include an overview, followed by presentations by each ISIC of technological and application progress.
Charles Romine, DOE program manager for the SciDAC applied mathematics centers, has organized a minisymposium on the subject for the SIAM 50th Anniversary Meeting.
Lori Freitag is a computer scientist in the Mathematics and Computer Science Division of Argonne National Laboratory.