ITaP and Purdue Calumet partnership

Todd Kalil, high-performance computing specialist for Purdue Calumet's Information Services Division, inside the Miner cluster.



Todd Kalil, high-performance computing specialist for Purdue Calumet’s Information Services Division, inside the Miner cluster.

At Purdue University Calumet, Chenn Qian Zhou’s engineering students work on real-world problems from energy companies, the steel industry, refineries and vehicle parts manufacturers, among others — businesses around Indiana, the nation or the globe.

These days, they are working a lot faster when running, say, complex fluid dynamics simulations, thanks to a supercomputer ITaP decommissioned in 2008 to make room for the Steele supercomputing cluster on Purdue’s West Lafayette campus.

Nodes from the former Lear cluster have been reborn — with the addition of a high-performance storage array and some other hardware by Purdue Calumet — as the Miner cluster. It is Calumet’s first large-scale high-performance computing system.

“Cases that we would run in a week we can run in less than a day,” says Tom Roese, a master’s student in mechanical engineering who is working with Zhou.

Zhou, head of the Mechanical Engineering Department, has purchased 100 licenses for FLUENT, a popular fluid dynamics software package, since Miner went online, so great has been demand from her students.

“As soon as the faculty and students understood the resource they had with Miner and the competitive advantage they could acquire with high-performance computing experience, they vigorously jumped at the opportunity to use the machine,” says Todd Kalil, high-performance computing specialist for Purdue Calumet’s Information Services Division.

Miner is a learning tool for Purdue Calumet students in addition to its role in faculty projects, which often aim to promote business retention and economic development in Northwest Indiana. High-performance computing experience has become a sought-after skill in industry, finance and even anti-terrorism and homeland security.

“Our objective is to enhance the learning experience for our students,” Kalil says. “When they go out in the world, they’ve got high-performance computing experience on a very large machine.”

Samuel Liles, a Purdue Calumet computer information technology professor, has built a senior level high-performance and distributed computing class around Miner.

“One of my students already is interviewing based on his work on the cluster,” Liles says. “The students are really jazzed over it.”

Professor Charles Winer, head of the Computer Information Technology and Graphics Department, says Miner is a good learning experience for his students, too.

“It’s different than anything else we have,” Winer says. “It enables us to do things that if we just worked on a desktop would take hours as opposed to minutes.”

Likewise, faculty members are now looking at problems they’ve never been able to before, Kalil says. The Northwest Indiana Computational Grid, a partnership including Purdue’s Calumet and West Lafayette campuses and Notre Dame, kicked off Miner with a grant program designed to spark groundbreaking research with demonstrable economic benefits.

The cluster also supports Purdue Calumet’s new Center for Innovation through Visualization and Simulation (CiVS), a data visualization lab aimed at student training and at economic development. Zhou, director of CiVS, is working to integrate its virtual reality and simulation capabilities with computational fluid dynamics and high-performance computing to enable interactive virtual design and engineering.

Her students examine issues like improving energy efficiency, meeting environmental requirements, optimizing production processes and enhancing product quality. One such effort, a senior project from student Dezhi Zheng, helped power company NIPSCO save $1.9 million by adjusting turning vanes to improve flow through ducts at one of its generating units.

Miner is being used for a variety of other purposes, from looking at the big picture — automated image analysis for disaster assessment and recovery planning — to the small — studying the pathways formed by molecules to facilitate biological functions.

Lear was a TOP500 supercomputer and its nodes are still quite capable, says Bill Whitson, director of research support at ITaP’s Rosen Center for Advanced Computing. Previously, Purdue Calumet’s largest machine was a 32-node cluster. Miner gives the campus a cluster with 512 nodes and 1,024 processing cores. Calumet purchased 30 terabytes of high-speed storage for the system.

Kalil says the Rosen Center worked closely with Purdue Calumet on getting Miner operational and continues to help with user and system support. The supercomputer is connected to West Lafayette via I-Light, the state fiber optic network that links Purdue’s campuses and other Indiana schools. The connection makes Miner part of DiaGrid, the Purdue-led multi-campus partnership that pools nearly 37,000 processors in student labs, offices, server rooms and high-performance clusters and puts them to work on research jobs when they would otherwise be idle.

Like the Steele, Coates and Rossmann clusters on the West Lafayette campus, Miner is named for a Purdue computing pioneer. Walter Miner was director of computing at Purdue Calumet and instrumental in creating the Calumet Computing Center.

Writer: Greg Kline, science and technology writer, Information Technology at Purdue (ITaP), 765-494-8167