As a research field, computational science and engineering (CSE) is relatively young, an integration of knowledge in computer science and computer-intensive science and engineering fields to help researchers put high-performance computers to better use. As an academic subject, CSE is even younger, yet CRPC researchers at the University of Tennessee, Rice University, and Syracuse University have helped make significant strides in CSE education, training the computational scientists and engineers needed to advance the use and development of CSE.
At the University of Tennessee, CRPC researcher Jack Dongarra and other faculty members formed a computational science program about four years ago. According to Bob Ward, professor and chair of UT's computer science department, "It was created in response to the desire to have a flexible program for meeting student needs [in CSE]."
Currently, there are 14 graduate and two undergraduate students involved in the program. Although there is no formal degree or certificate program, students receive a thorough introduction to CSE through related research and classes. Computer science coursework covers computer systems organization, performance evaluation, linear systems, computing applications, numerical methods, and parallel and distributed computing.
Regarding research, the program has maintained student interest in CSE by involving individuals from different fields and organizations. Along with 10 faculty members from the UT computer science department, the program involves collaborators from UT's mathematics and engineering science and mechanics departments and its medical center. It also draws upon the expertise of collaborators from Bell Laboratories, Cray Research, Inc., Argonne National Laboratory, Oak Ridge National Laboratory, the U.S. Department of Agriculture, and other universities. By involving many partners from government, industry, and academia, the program exposes students to CSE from a broad range of perspectives.
Research also allows students to play significant roles in major projects. Several CRPC-related projects have ties to the program, including ScaLAPACK, PVM, MPI, and the National HPCC Software Exchange. "Good examples abound for significant student involvement in research," says Ward. He specifically mentioned the work of one student, Qingping Deng, who is pursuing an M.S. in computer science and a Ph.D. in mathematics. Deng is currently studying parallel adaptive mesh techniques for fluid problems in preparation for collaborative research with computer scientists at Argonne National Laboratory. Dongarra mentioned the work of Antoine Petitet, a Ph.D. student in computer science who was instrumental in the design and implementation of Parallel Basic Linear Algebra Subprograms (PBLAS).
Besides providing a rich background for students, partnership with outside collaborators fulfills many of the program's goals: knowledge exchange within the CSE community, dissemination of CSE educational methods to other universities, promotion of the use of computationally intensive tools, and advancement of computer technology. As an effort dedicated to both education and research, UT's CSE program is advancing the field in an integrated, hands-on fashion.
Established in 1992, Rice's Computational Science and Engineering Graduate Degree Program is one of the first programs of its kind in the country. The CSE program offers professional master's and Ph.D. degrees that prepare candidates to work as computational specialists in either chemical engineering, applied mathematics, biology, electrical engineering, statistics, or mechanical engineering.
"We realized that because of the boost in computing power available to scientists and engineers, computation has increasingly become an integral part in analysis, development, and design," noted CRPC professor and researcher Dan Sorensen, who coordinates the program among the computer science department and several other science and engineering departments at Rice. "There is really a need for scientific computing experts who could work as part of an interdisciplinary research team. This program addresses that need."
Students learn to apply advanced hardware and software as well as new algorithms and analytical techniques useful to their research field. Classes cover subjects like optimization methods, concurrent programming, linear and nonlinear programming, and algorithms for linear algebra and differential equations.
Like UT's program, the Rice CSE program involves researchers from multiple disciplines and institutions, such as Caltech, the University of Illinois, and Stanford University. Several students have been involved in CRPC-related projects at Rice. The program has even produced its first Ph.D. graduate, Richard LeHoucq, who was awarded the 1995 J. H. Wilkinson Fellowship in Scientific Computing at Argonne National Laboratory.
For the past three years, Syracuse University has offered an undergraduate and graduate program in computational science. The program draws on many existing courses in applied math, computational techniques for scientific and engineering areas, and computer science, especially those in high-performance computing. Faculty from several departments at Syracuse have combined the study of computer science with specific areas of engineering or science.
At the undergraduate level, students can receive a "Minor in Computational Science." Graduate students can either receive a "Certificate in Computational Science," along with a traditional master's or Ph.D., or the recently established "Master's Degree in Computational Science."
Students are introduced to the CSE program through an "Introduction to Computational Science" course, which contains modules covering typical application areas in particle systems, field simulations, statistical techniques and optimization problems. Each module covers all aspects of a problem, from the numerical methods and computational algorithms to computing issues for those algorithms and their use in specific problems. Other courses give practice in sophisticated, field-level techniques and are supported by the computing resources of the Northeast Parallel Architectures Center (NPAC).
"The new computational science courses have been very successful," says NPAC research scientist Nancy McCracken, who serves as chair of the computational science curriculum committee. "About 30 students complete the introductory course and 15 go on to the advanced projects course each year. The program has been particularly strong in student projects in Physics and Computational Fluid Dynamics."
In addition to the typical scientific and engineering projects, more recent developments have been in areas related to high-performance computing techniques for information-based applications. This has been possible at Syracuse because of its strong professional schools in broadcast journalism, information studies, and public policy. Computational science curriculum is also being developed in these disciplines.
Like other CRPC-related outreach efforts, these programs emphasize the "people using the machines." Besides preparing students, CSE programs at the University of Tennessee, Rice, and Syracuse serve as models for other universities interested in establishing their own CSE programs. Already, several universities have followed suit. CSE programs, like many CRPC-related efforts, aim to benefit the entire HPCC community.