TABLE OF CONTENTS FRONT PAGE NEXT ARTICLE PREVIOUS ARTICLE
Like planes waiting to land at the airport, critical data from faculty and
student research projects line up and wait their turn to be processed through
the University's central computing station in Phillips Hall.
But unlike the airport, the wait is minor with 60 processors and the
capacity to handle 34 billion bytes of information in computer memory at any
given time.
That's the way Judd Knott, director of Academic Computing Systems in
Academic Technology and Networks (ATN), described his office's ability to
support University researchers and deliver efficient, cost-effective
results.
"Researchers want to get answers quickly," Knott said. "Faculty and
students who don't have the application or local resources they need, who are
limited by their department desktop computers, who have large volumes of
mission-critical data that need to be stored: Those are the people who benefit
the most from our services."
With servers busy round the clock and running at capacity 365 days a year,
the economies of not having computers stand idle are paying off in faster
results at a lower cost.
An additional server, purchased in cooperation with the chemistry
department, helped expand service to the entire University.
"When it's not being used for their own research projects, other campus
research is automatically queued into their server and runs until they need it
again," Knott said. "This pooling of resources reduces our costs
proportionately."
Reorganization improves access, development and efficiency
As a result of a recent reorganization of ATN, five professional
staff and five graduate students are on-site to support research computing
services in five major areas under the management of Ruth Marinshaw. The five
areas are:
* Bioscience computing, which provides gene- and protein-sequence analysis
tools to more than 600 researchers on campus, as well as courses and help for
researchers in conjunction with the Center for Bioinformatics.
* Statistical computing, which provides researchers with a wide selection
of statistical analysis programs and hosts more than 1,000 users.
* Scientific computing, which provides for computer-intensive applications
not strictly statistical in nature, including computational chemistry, all
types of simulations, molecular modeling, geographical information systems,
physics and data visualization.
* Data management services, which route and schedule projects to different
machines and enable access to more than 120 software packages.
* Distributed computing infrastructure (DCI), which provides a technology
infrastructure to the University community. It enables users to plug into and
access a host of instructional resources such as course content databases and
computational software programs. University students, faculty and researchers
can obtain an ATN UserID for free on the http://help.unc.edu web site.
Undergraduate students need a faculty sponsor to subscribe.
Keeping faculty focused on research
Sixteen campuses have access to the N.C. Super Computing Center in
Research Triangle Park, but it's the important research computing services
available right here on campus that set Carolina apart, according to John
Oberlin, ATN executive director.
With specialists in each of the five service areas, big memory, large
storage, parallel processing and a commitment to Carolina faculty, the
University is well-equipped to handle the rigorous demands of campus
researchers.
"Our goal is to provide resources, training and direct support to our
faculty so that instead of worrying about the computing, they can put full
energy into their research discipline," Oberlin said.
"We're committed to helping them find answers that make sense, to make the
process painless and efficient and at the same time cost-effective to the
University."
And ATN is succeeding, according to Donna Gilleskie, assistant professor
of economics.
"Access to a fast and large platform for statistical computing, like the
ATN platform, has been essential for carrying out my research in health
economics," she said.
"For example, I am in the process of testing a new estimation procedure
that requires running thousands of programs, each with small modifications to
several control variables. I simply submit the programs in batch to a queue and
let them run as CPU time becomes available."
To UNC-CH Home Page