Illinois Campus Cluster Program offers access to high-performance computing

With the startup of the Blue Waters petascale supercomputer at Illinois, much of the attention in high-performance computing has focused on ways to harness this new level of computing power and speed. Although Blue Waters is currently among the fastest and most powerful supercomputers in the world, it is just the latest entry on a long list of computing innovations developed at the University of Illinois. Illinois has additional high-performance computing resources available to researchers.

“High-performance computers have become essential to research in many disciplines across the Illinois campus, from Big Data applications, to astronomy, biology, and chemistry on through the alphabet to zoology,” explained Chuck Thompson, assistant dean and chief information officer for the College of Engineering. “The Illinois Campus Cluster Program is a campus-wide resource that helps meet this need for research computing cycles.”

“This is a computing campus. If you go around the country, that’s how others refer to us,” explained Narayana Aluru, director of the Computational Science and Engineering (CSE) program, which manages a significant investment in and access to a large portion of the current Campus Cluster.

“The Campus Cluster is a vital tool that supports engineering faculty research, student research, and classroom education here at Illinois,” Aluru added, noting that CSE is a cross-cutting, interdisciplinary unit with more than 200 faculty affiliates across campus. CSE is not only providing access to the Campus Cluster, but it also provides training and support for CSE-affiliated users. (NCSA handles hardware installation, maintenance and baseline support of the systems). 

As a major player in ‘Big Data,’ CSE is an enabler for great science and engineering.

“My research group runs a wide range of applications on the Campus Cluster, from the multi-scale finite element modeling of damage evolution in thin metallic films, to the molecular dynamics modeling of the spallation failure of self-assembled monolayers,” explained Philippe Geubelle, a Bliss Professor and head of the Department of Aerospace Engineering. “Access to the Cluster has been really instrumental in our ability to tackle larger, more complex problems.”

“We’re doing something significant and unique—not available on other campuses,” Aluru said. “CSE is a recognized discipline, recognized internally as a pillar of innovation.” CSE currently supports 550 user accounts whom collectively used 28.7 million CPU hours last year. The next largest group behind CSE used 3.6 million hours.

“We’ve had some excellent results…a number of papers in prominent journals," Aluru added.

The concept of a Campus Cluster was developed about 15 years ago by former CSE director Michael Heath in 1999. The Turing Cluster was the first high-performance cluster with wide availability to the campus community. It served graduate and undergraduate students doing computing research as well as faculty research across the College of Engineering and campus for about 10 years.

Ten years later, Heath, a professor and a Fulton Watson Copp Chair in the Department of Computer Science, talked to a number of units who pooled their money to help purchase much of the hardware for the first instance of the Illinois Campus Cluster, Taub, which entered production mode in May 2011.

The Taub Cluster is named in honor of Abraham Haskel Taub (1911-1999), chief mathematician for the original ILLIAC project and director of the Digital Computer Laboratory from 1961 to 1964. The plan for the Illinois Campus Cluster is to build and sustain several clusters in operation at once. The new Golub Cluster will be fully operational in early fall 2013.

Individuals, groups, and campus units can invest in compute and storage resources on the clusters, which are housed at the Advanced Computation Building (1011 W. Springfield Ave., Urbana) a facility specially designed to support high-performance computing systems. Those who invest in compute resources on the cluster have guaranteed access to the number and type of nodes in which they invested.

“Any researcher at Illinois can contribute and any contributor can benefit,” said Aluru, who is the Kritzer Distinguished Professor in the Department of Mechanical Science and Engineering. “When any owner is not using his or her share, any other user may take advantage of the surplus,” Aluru said. “In this way, researchers get more computing capacity for their money than they would by building many separate systems across campus.”

Pooling computer resources also helps researchers control expenses and reduce overhead, shorten startup time, avoid space renovation and free up space formerly used for separate cluster resources, and reduce the campus carbon footprint. Students can get accounts for courses in high-performance computing as well as courses that take advantage of high-performance computing.

Computational Science and Engineering is a relatively new paradigm for scientific research and engineering design in which large-scale simulation, data analysis, and high performance computing play a central role. CSE is inherently interdisciplinary, requiring expertise in advanced computing technology as well as in one or more applied disciplines.

For more information, see http://campuscluster.illinois.edu or contact help@campuscluster.illinois.edu.
_____________________________

If you have any questions about the College of Engineering, or other story ideas, contact Rick Kubetz, writer/editor, Engineering Communications Office, University of Illinois at Urbana-Champaign, 217/244-7716.