At Qumulo Qonnect East–a live virtual event, Ben Gitenstein, VP of products, spoke with Eric Tomeo, a system administrator in the research computing department at the University of Florida (UF). The university has been using Qumulo’s file system for over 5 years.
“We manage the supercomputer,” said Tomeo. “We have 60K compute cores, about 600 GPUs and this year we added an NVIDIA DGX SuperPod.” The UF’s HiPerGator 3 supercomputer integrated 140 NVIDIA DGX A100 systems powered by a combined 1,120 NVIDIA A100 Tensor Core GPUs to enable the rapid scaling of massive AI infrastructure. “This was a major and welcome addition; and, quite the system to manage,” said Tomeo. It includes 4 petabytes of high-performance storage, running on Qumulo’s file data platform.
“With the addition of the SuperPod, we landed #22 of the top 500 supercomputers in the world,” said Tomeo. “We are very proud of that accomplishment.” Gitenstein commented that the researchers must be in their element too!
Watch the short YouTube video to listen in on the rest of the conversation.
How the UF manages data in a high-performance research computing environment
According to Tomeo, UF research spans all areas of biotech including geo sequencing, RNA, DNA modeling, machine learning, deep learning, and AI.
When asked why his team chose the Qumulo file system for research computing, Tomeo said, “We wanted a system that could stand up and scale in the supercomputing environment. Qumulo’s licensing was simple and all inclusive; unlike other file system providers whose licenses charged for everything. We wanted something modern, and that’s why we chose Qumulo.” explained Tomeo.
It was a good choice. Any problem that the supercomputer center had before was remedied by the system, “the dashboard analytics are great, performance is great, upgrades are a no brainer,” Tomeo said. “Originally we didn’t think upgrades were going to be so easy; all we do is put the file in place and click the upgrade button.”
“Qumulo does containerized updates so fast, nothing even notices. When we put VMs on, there are some time outs with highly available systems, but most updates are hitless, it’s nondisruptive.”
Watch the rest of the interview to learn more about how the research computing team at UF manages data at the supercomputer center.
Related resources
- Real-time analytics use case: How the University of Florida gains insights into research computing data with Qumulo
- Solution page: See why top research computing institutions across the world rely on Qumulo to provide high performance data capture with a single platform that unifies analytics, data process and storage
- The Qumulo difference: See what smarter data management means for you — and how Qumulo delivers the same experience across on-premises and cloud environments
