UpDate - Vol. 14, No. 26, Page 9
April 6, 1995
TechTalk
Gershwin joins composers on UD network

     The number of composers on the University's network increased
recently with the debut of Gershwin, a Silicon Graphics Power
Challenge that will advance high-level research across a range of
scientific and  engineering disciplines.
     The Power Challenge, with eight CPUs and two gigabytes of memory,
will be used by researchers to do high-performance computing. The
system has 40 gigabytes of disk space.
     The Power Challenge's purchase is the result of a plan submitted
to the provost and deans by the University's Large-Scale Computing
Advisory Committee for improving research capabilities and access to
external funding at the University. The plan also includes the
purchase of a Cray J90 supercomputer that is to arrive this summer.
     The computing power these systems offer will greatly enhance
external funding opportunities for University researchers as they
compete in a shrinking research-dollar marketplace.
     "The Power Challenge system's industry-leading, shared-memory
multiprocessing technology will allow University of Delaware
researchers to solve highly complex scientific and engineering
problems more efficiently and intuitively," Ron Bernal, vice president
and general manager of Silicon Graphics Supercomputing Systems
Division, said. "And, in addition to realizing high-level research
objectives, the University will expose students to the latest
innovations in high-performance computing and parallel processing."
     "Our high-end research community had a collective need for a
powerful new-generation supercomputing environment; at the same time,
however, cost-effectiveness was a major consideration," according to
Daniel J. Grim, chair of the Large-Scale Computing Advisory Committee
and executive director of network and systems services.
     "The Power Challenge system met our exacting price/performance
requirements. And, its excellent performance in both multiprocessing
and parallel applications and ease of programming also played a major
role in the selection process," Grim said.
     Science and engineering faculty will use Gershwin to further
their research and teaching. The system also will be used for
instructional purposes, including the teaching of courses in high-
performance and parallel computing.
     "For what I want, Gershwin came at the right time, and I can use
it for many interesting, high-performance computer simulations," said
Lian-Ping Wang, mechanical engineering.
     Wang uses supercomputers to explore the physics of turbulence and
two-phase flows and to improve the modeling aspects of these flows.
     According to Wang, this is only possible if there is a large
degree of freedom to work with and use an efficient numerical scheme.
"The speed of the supercomputer is critical to a reasonable turnaround
of a production run where one has to advance the system in small
increments of time over a long time," Wang said.
     Victor Kaliakin, civil engineering, said he plans to use Gershwin
for research in computational mechanics. Kaliakin works on simulating
soils and structures of various types subjected to different forms of
loading.
     "The Power Challenge machine will likely provide the increased
computational speed so that large problems can be solved in a
reasonable time," Kaliakin said.
     Eric Mas, a fourth-year graduate student in the physics
department, said he looks forward to using Gershwin for research
purposes. "My group analyzes molecular systems from first-principles
using quantum mechanics, but our ability to quantitatively describe
interaction between two even very small molecules using only first-
principles theory is limited."
     Mas and his group plan to use Gershwin to model molecules and to
predict the outcome when the molecules interact. According to Mas, all
this work will require very powerful computers.
     "By being able to predict intermolecular interactions on the
computer, scientists will be in a position to design molecules with
desirable properties, which will have obvious medicinal applications,"
Mas said.
     "We will be able to model proteins, enzymes and eventually
strings of DNA."
     "Of course, we are in the infant stages of development, but the
future looks pretty exciting, and Gershwin's speed and memory capacity
will allow us to take a few more steps forward," Mas added.
     If you are interested in an account on Gershwin, send e-mail to
Jack Seltzer in user services (jack@strauss.udel.edu).