ALCF summer students gain experience with high-performance computing

Jim Collins

Facebook Twitter LinkedIn Google E-mail Printer-friendly version

With an internship at the Argonne Leadership Computing Facility (ALCF) this summer, Elmhurst College senior Regan Brown got a unique opportunity to work at the forefront of high-performance computing.

“I knew there was a lot of new and exciting stuff going on in the world of computing that my textbooks couldn’t keep up with,” said Brown, who is majoring in computer science and computer game design. “Getting exposed to many-core machines and parallel programming at Argonne sounded a lot more appealing than working with outdated technology at non-scientific firms.”

Brown's internship, through Science Undergraduate Laboratory Internship (SULI) program supported by the DOE's Office of Science, paired her with mentor Ray Loy, an ALCF application performance engineer, for a project that involved researching the performance of debuggers and how they can be optimized for Mira, the fifth most powerful supercomputer in the world.

“Getting to work with one of the fastest supercomputers in the world is a once-in-a-lifetime experience that not many people get to have,” Brown said. “It gave me insight into what I can expect to happen in computer science research and the computing industry in the next few years.”

Brown was part of the ALCF’s largest summer student class to date – a group of 24 students who ranged from college freshmen to PhD candidates. This year’s crop of students tackled a wide variety of projects that covered everything from system administration and data analytics to computational science and performance engineering.

“We have students working with things like RAM disks and data visualization, which can be very attractive to research organizations as well as the big tech companies,” said ALCF Director Michael E. Papka, who is also an associate professor of computer science at Northern Illinois University.

Every year, the ALCF, a DOE Office of Science User Facility, solicits student project proposals from staff members who are interested in mentoring. Through programs like SULI and Argonne’s Research Aide Appointments, the students are brought in to work alongside their mentors on real-world research projects. The program culminates with a special symposium in which the students present their project results.

“It’s a win-win situation,” Papka said. “The students gain valuable experience collaborating with researchers in a national laboratory environment. We get to support education and we benefit from the insights they uncover along the way.”

For one project, Judah Unmuth-Yockey, a PhD candidate at the University of Iowa, worked with his mentor James Osborn, an ALCF computational scientist, to develop parallel code for the Tensor Renormalization Group (TRG) method, a relatively new approach to solving systems on a lattice that can, in principle, be applied to a wide range of physics problems.

“My ultimate goal, like many others, is to solve outstanding problems in nuclear theory, high energy physics, and quantum gravity,” said Unmuth-Yockey, who came to Argonne for a second consecutive year through the research aide program. “Working on the lattice and studying emergent structure and effective theories at the ALCF seems like a good step towards being able to work on those problems effectively.”

In addition to building on his PhD research with numerical renormalization group techniques, Unmuth-Yockey said the position gave him experience programming with Python and C/C++ and an opportunity to make connections with other young students from around the country and the world.

Another summer research aide, Amina Hussein, came to the ALCF for a chance to apply her computing skills in a new and different way.

The recent Purdue University graduate has a master’s degree in nuclear engineering with a specialization in computational physics. But rather than working on a computational science project, Hussein ventured outside of her comfort zone to work with a performance analysis, projection, and tuning tool called SKOPE (SKeleton framewOrk for Performance Evaluation).

With help from her mentor Vitali Morozov, ALCF principal application performance engineer, Hussein validated and enhanced SKOPE’s compiler to improve its ability to model and explore an application’s workload. The changes will result in more accurate performance studies on current and future supercomputers.

She believes the hands-on research project and the opportunity to collaborate with leading experts in the field will serve her well in the future.

“Through this project, I’ve gained valuable knowledge on performance enhancement, programming abstraction, and the transformation of high-level to low-level code,” Hussein said. “This experience has augmented my perspective as a computational scientist, and I look forward to applying these skills as a doctoral student at the University of Michigan this fall.”

And that’s exactly the outcome Papka hopes for with the ALCF’s summer student program.

 “If we give the students an experience that can reshape how they approach their future, then the program is a success,” he said. “We hope this puts them on a path where they eventually come back and work at DOE national laboratory, but we’re also content knowing this experience could lead them to bring ideas of high-performance computing into the business world.”

Regan Brown’s SULI position was funded through the DOE Office of Science’s Office of Workforce Development for Teachers and Scientists. Students from Argonne’s Research Aide Appointments and the Student Research Participation Program were supported by the ALCF and DOE’s Advanced Scientific Computing Research program.

2015 ALCF Summer Student Projects

Alex Ballmer
Education: BS (Spring 2018), Computer Science, Illinois Institute of Technology
Mentor: William Scullin
Project: Automating the Build Process of HPC Applications

Regan Brown
Education: BS (Spring 2016) Computer Science, Elmhurst College
Mentor: Ray Loy
Project: Parallel Debugging on Blue Gene/Q: A Performance Perspective

Johnny Bui
Education: BS (Spring 2018), Computer Engineering, University of Illinois at Chicago
Mentor: Hal Finkel
Project: Systematic Development of a Loop Benchmark Suite for the Evaluation of Current and Future HPC Hardware and Compilers

Cameron Christensen
Education: MS (Fall 2015), Computing – Graphics & Visualization, University of Utah
Mentor: Venkat Vishwanath
Project: Efficacy of Cloud‐Based Data‐Centric Frameworks for HPC Data Analysis

Alex Findlater
Education: PhD (Spring 2016), Physical Chemistry, Iowa State University
Mentors: Yuri Alekseev, Graham Fletcher, and Spencer Pruitt
Project:  Performance of Matrix Operations in GAMESS

Barry Greengus
Education: BS (Spring 2017) Computer Science, University of Illinois at Chicago
Mentor: Bill Allcock
Project: A Performance Analysis of a RAM Area Network

Amina Hussein
Education: MS (Spring 2015) Nuclear Engineering, Purdue University
Mentor: Vitali Morozov
Project: Automatic Skeleton Generation for Scientific Codes

Jie Jiang
Education: PhD (Fall 2017), Industrial Engineering, University of Illinois at Chicago
Mentor: Silvio Rizzi
Project: Exploration of Multiple Streaming from vl3 to Tiled Display

Kristopher Keipert
Education: PhD (Fall 2017), Physical Chemistry, Iowa State University
Mentors: Yuri Alekseev, Graham Fletcher, and Spencer Pruitt
Project: Performance of Matrix Operations in GAMESS

Michael Lewis
Education: PhD (Spring 2016) Computer Science, University of Illinois at Chicago
Mentor: Preeti Malakar
Project: Optimizing Read Times for Multidimensional Datasets in S3D Application

Xiayue (Patrick) Li
Education: BS (Spring 2017), Physics, Math and English, Tulane University
Mentor: Alvaro Vazquez‐Mayagoitia
Project: Predicting Molecular Crystal Structure with High‐Throughput Design of Genetic Algorithm on Mira

Phil Lindner
Education: MS (Fall 2016), Computer Science, Northern Illinois University
Mentors: Michael E. Papka and Venkat Vishwanath
Project: Investigating Event Handling Mechanisms and Control Stream Metadata in the Context of M‐to‐N Network Data Transfers

Amrut Marathe
Education: MS (Fall 2015), Computer Science, Northern Illinois University
Mentor: Doug Waldron
Project: Analysis of the Node Failures in High‐Performance Computing Systems

Santoshi Nanagare
Education: MS (Summer 2015), Computer Science, Northern Illinois University
Mentor: Doug Waldron
Project: Application Development for ALCF’s User Experience Team

Adolfo Rodriguez
Education: MS (Winter 2015), Computer Science, Northern Illinois University
Mentor: Tom Uram
Project: Understanding Machine Learning with High Energy Physics Data

Min Shih
Education: PhD (Summer 2016), Computer Science, University of California at Davis
Mentor: Joe Insley
Project: Advanced Rendering in vl3

Luis Antonio Soriano Agueda
Education: PhD (2017), Chemistry, Universidad Autonoma Metropolitana-Iztapalapa
Mentor: Alvaro Vazqez-Mayagoitia
Project:  NWChem/SEMO

Joe Sortino
Education: BS (Spring 2015), Computer Science, North Central College
Mentors: Tom Uram
Project: Volume Rendering as a Service

Judah Unmuth‐Yockey
Education: PhD (Spring 2016), Lattice Field Theory, University of Iowa
Mentor: James Osborn
Project: TRG Method for the O(2) Spin Model

Ben Walters
Education: BS (Spring 2016) Computer Science, Illinois Institute of Technology
Mentor: Bill Allcock
Project: A Performance Analysis of a RAM Area Network

Roger Xiao
Education: BS (Spring 2016), Computer Engineering, University of Illinois at Urbana‐Champaign
Mentor: Kevin Harms
Project: Evaluation of IBM exFLASH Memory

Adam Young
Education: MS (Summer 2015) Computer Science, Northern Illinois University
Mentor: Michael E. Papka
Project: Holistic Visualization of High‐Performance Computing Facilities

Xingwu Zheng
Education: PhD (Summer 2016), Electrical Engineering, Illinois Institute of Technology
Mentor: Vitali Morozov
Project: Exploring the Opportunities for Data Staging in HPC Workloads

Zhou Zhou
Education: PhD (Winter 2015), Computer Science, Illinois Institute of Technology
Mentor: Paul Rich
Project: Planning and Optimization in Cobalt Resource Manager