Universities Work to Optimize High-Performance Computing Resources
For higher education institutions, high-performance computing is essential to enabling academic and scientific research arms to support and improve pathways for their researchers.
To help uncover issues in their networks and maintain optimal performance, universities are starting to ask the right questions: Who is using the network? For what? How much bandwidth are projects consuming?
SIGN UP: Get more news from the EdTech newsletter in your inbox every two weeks!
How to Design for Optimal Performance for Research
For universities looking to utilize their research findings, the first step is understanding the weaknesses in their HPC network.
The University at Buffalo, with funding from the National Science Foundation, developed software to monitor HPC resources to help with optimization.
XD Metrics on Demand (XDMoD) runs in all NSF-funded HPC centers, where it collects data on and reports to NSF about utilization and job-level performance, explains Thomas Furlani, Director of the university’s Center for Computational Research. There’s also an open-source version used by a few hundred academic and commercial HPC centers worldwide.
In addition to gathering data such as CPU capacity, disc I/O rate and cache, XDMoD runs application kernels daily to measure quality of service.
“Since we run them every day, when something gets out of sync, we can tell that the performance has dropped and ask what happened between then and now to cause the system performance to be poor,” Furlani says. “We don’t want to wait for users to notice problems. This allows us to find them before the canaries in the coal mine.”
Strong University HPC Network Can Extend Data's Reach
Even with a strong network, most university research holdings tend to be treated as data islands. There’s no central inventory of projects, so data rarely extends beyond its primary use.
Montana State University set out to change that, using its Bridger High Performance Research Network to conduct a data census with the goal of turning their data islands into lakes.
In one project, Ross Snider, an electrical and computer engineering researcher at MSU, is working with colleagues at the University of California, San Diego to study marmosets’ vocalizations. The researchers use BridgerNet to share the hundreds of gigabytes of data they capture.
By studying how marmosets use discrete signals to communicate, the researchers hope to create algorithms that can detect audio signals in order to improve the next generation of human hearing aids.
When MSU Vice President and Chief Information Officer Jerry Sheehan and his team learned of the project, they connected Snider with opportunities for university outreach to K–12 students.
A Family Science Night event at the university gave 500 students the opportunity to learn about Snider’s research by viewing content from thermal-imaging cameras and audio-imaging software depicting the marmoset vocalizations.
“That opportunity to engage and inform doesn’t happen if we don’t know about Ross’s data,” says Sheehan. “That fusion of research, instruction and outreach doesn’t happen if we can’t bring together programs.”
To learn more about high-performance research networks, read Big Data Networks Connect Higher Education Researchers.
