Spurring Scientific Exploration
Spurring Scientific Exploration on the Great Plains with XSEDE and HPC
By Scott Gibson
Doug Jennewein, research computing manager at the University of South Dakota. [Image courtesy: the University of South Dakota]
Computer scientist Doug Jennewein's working environment is, in more ways than one, a great frontier. Situated at the University of South Dakota (USD) on America's wind-swept plain, he is in a job role that is allowing him to pursue his passion and be a pioneer in growing the use of computationally assisted research.
"I've always been interested in science and technology, usually as separate fields, but working in research computing has provided the opportunity to scratch both itches, so to speak, which is very exciting," says Jennewein, USD's research computing manager.
When he was an undergraduate at USD, a friend's talk of building a "Beowulf cluster" intrigued him, and later as a research analyst in the computer science department in 2004, Jennewein deployed a proof-of-concept cluster, which in turn led to the department's securing funding for a small cluster that was brought online in 2006 and expanded in 2008, 2009, and 2012 with grant and institutional support.
Just beginning his job as research computing manager in 2012, he was looking for lots of advanced computing training materials, especially what might be available online. High-performance computing center management and growth were of interest to him, but even more, he wanted to learn about science engagement, parallel programming, and file system I/O, as well as specific programming languages such as R and Fortran, which are common in the research setting but not so much in mainstream computing.
He recalled having heard in 2011 through the Great Plains Network (a consortium of universities in the Midwest) newsletter about the transition of TeraGrid to XSEDE and a discussion of the new XSEDE portal. That portal turned out to be the entryway to exactly the type of training he was after.
"There's just a vast variety of topics that has been covered by XSEDE, and that's been very valuable to me as a relatively early career professional in a remote rural state," Jennewein says.
He approached XSEDE's online webinars to learn about unfamiliar topics, such as certain programming languages, but also for enlightenment concerning best practices for conducting training and outreach. He wanted to observe how XSEDE conducts a class and how it presents complicated subjects to a new audience. "I did this for my own benefit and to improve my own training and outreach here on campus," he says.
In addition to what to he's gleaned from the webinars, he's gained insight from the cluster user guides and application user guides. "Both told me about the resources at XSEDE but also showed me kind of a large-scale reference implementation of how to organize and present user guides and documentation," he explains.
Jennewein's current career objectives are to engage USD researchers and determine ways he can support them in leveraging HPC in their projects, while in the process furthering his knowledge of domain sciences.
While USD's computing profile has largely been small task-parallel jobs, with limited use of MPI and distributed memory, Jennewein aspires to leverage his position to foster the growth of advanced computing at the university.
One of his strategies is what he refers to as "research computing house calls," which entail visiting with researchers and their team in person to determine their needs and to demonstrate interest in and support of their respective scientific domains. The "research computing house calls" can be initiated through his response to a help desk request or by attending a departmental seminar or workshop.
Currently, the largest community involved in computationally assisted research at USD is high-energy physics, but user groups from the materials chemistry and bioinformatics disciplines also need advanced computing support. The bioinformatics community constitutes a prime set of current and prospective users because USD is home to the Sanford School of Medicine.
"As new faculty come to campus, they expect computational resources, especially bioinformatics folks involved in high-throughput sequences or next-generation sequencing," Jennewein says. "Particularly for those use cases, we are looking at XSEDE as a scale-out option."
He adds that XSEDE makes itself an attractive choice to the USD bioinformatics researchers and others by affording them an on-campus liaison (Jennewein) for assistance in writing a compute allocation proposal to gain access to national-level advanced digital resources.
Recently, Jennewein discovered a way to further solidify his relationship with XSEDE by becoming a Campus Champion (CC). Working with CC Program Coordinator Kay Hunt of Purdue University, Jennewein recently took the initial steps to join the program.
Rolling out the Champions program at USD is one of Jennewein's main goals for 2015. "I definitely look forward to that," he says. "Also, further interactions with the other Champions, as well as attending the XSEDE15 conference. That will be a new experience for me." And he will be equipped with more knowledge to share during "research computing house calls" on the Great Plains.
Podcast