HPC Ignites Collaboration: How Missouri’s Hellbender System Creates Community

Meet the hellbender, an ancient and elusive salamander found in just a few river systems across the United States. Missouri is the only state where both of its subspecies live, making this remarkable creature especially significant to the region.

The hellbender is more than a biological curiosity. It is a bioindicator, meaning its well-being reflects the health of its entire environment. Clean water, stable habitat, and balanced ecosystems are essential for its survival. When the hellbender is thriving, it signals that the natural world around it is in harmony.

This deep connection between environment and vitality inspired the name of the University of Missouri’s newest high performance computing (HPC) system.

Like the salamander, the Hellbender HPC cluster represents a complex system where balance, collaboration, and access are key. It stands as a sign of a growing innovation ecosystem, one where diverse minds can come together and flourish.

The collaborative spirit behind the Hellbender ecosystem is exactly what SC25’s theme, HPC Ignites, aims to celebrate. As the conference approaches this fall in St. Louis, the theme invites us to consider what HPC truly sparks — both in research communities and around the world. For the University of Missouri (MU) System, one clear answer is an expanding impact that reaches both deep into disciplines and wide across institutions. “The biggest transformation has been the scale of research that’s now possible,” said Matthew Keeler, Director of IT Research Support Solutions at MU. 

The biggest transformation has been the scale of research that’s now possible.

— Matthew Keeler, Director of IT Research Support Solutions, MU

While Keeler focuses on research support strategy at the Columbia campus, approximately 100 miles away, in Rolla, Buddy Scharfenberg leads similar efforts at Missouri S&T. He notes that building and operating shared infrastructure across campuses is creating more than just technical capacity; it is fueling a new culture of collaboration. “Matthew had the researcher connections. I had the technical resources. Pairing those two is what made this possible. If Matthew hadn’t been up there beating the drum, saying, ‘Come check this out,’ we’d never have reached this level of utilization,” Scharfenberg said.

That spirit of collaboration between campuses, researchers, and technologists has become the defining feature of the Hellbender ecosystem.

HPC has traditionally been reserved for users with deep technical expertise — those comfortable with SSH terminals, job schedulers, and command-line interfaces. Hellbender is working to change that. By integrating Open OnDemand, a web-based platform developed by the Ohio Supercomputer Center, the system is opening HPC access to a much broader range of users.

Open OnDemand offers a graphical interface that enables users to run jobs, launch applications such as VS Code and Jupyter Notebooks, and manage files, all without requiring command-line knowledge. This shift is creating new entry points for researchers who may not have a background in computer science.

“The addition of Open OnDemand has brought several benefits,” explained Keeler. “First, it makes the system more accessible to users who aren’t familiar with command-line environments, which is especially helpful for beginners. Second, it lets us place constraints on how resources are used.” He noted that this approach also enables the team to route processes more efficiently, thereby improving system stability and enhancing the overall user experience. 

“We customized the OnDemand interface to look less intimidating and more like something they’re familiar with,” said Scharfenberg. “That simplification has been a key innovation we introduced with Hellbender, one that really addressed a major usability gap.”

John Harrison, Senior Advanced Computing Environment Manager, helps oversee the Hellbender HPC system in MU’s data center.

Allocating HPC resources has always been a balancing act between fairness and priority. Traditional approaches often force researchers to wait in long queues for access that isn’t guaranteed, or to invest in costly dedicated systems that can create silos and limit broader community benefits. Hellbender offers a smarter alternative by combining a “condo model” with a community-first philosophy that ensures resources are both reliable and shared efficiently.

“Historically at MU, even if you were an investor in the system, you still couldn’t guarantee immediate access,” explained Keeler. “What sets Hellbender apart is that we offer preemption rights. If you own resources, you can run your jobs when you need them. That’s a meaningful guarantee that supports time-sensitive research.”

Here’s how it works: when researchers invest in compute nodes, they become “investors” in the system. This grants them priority access to the hardware they funded. But rather than letting those resources sit idle when not in use, Hellbender automatically returns unused capacity to a shared pool, making it available to any researcher across the university.

Scharfenberg offers a helpful analogy: “Think of it less like owning a whole condo where no one else is allowed in, and more like having a reserved study room in a public library. If you’ve invested in it, you get priority access. You can book it whenever you need it, but when you’re not using it, someone else can. That’s the model we’ve built, and it makes resource use more efficient and the community more collaborative.”

Think of it less like owning a whole condo where no one else is allowed in, and more like having a reserved study room in a public library. If you’ve invested in it, you get priority access.

— Buddy Scharfenberg, Director of IT Research Support Solutions, Missouri S&T

Keeler adds that while the condo model itself isn’t unique to Missouri, their implementation takes it further. “It’s not a model that’s unique to us. It’s actually pretty common in the HPC world,” he said. “But what sets us apart is the shift from systems that only vary prioritization over time to one that guarantees committed access when needed.”

Several key design choices make Missouri’s version stand out. Half of all system resources are reserved for general campus use, and researchers only pay for hardware, not for administrative or support costs. This prevents the “haves versus have-nots” dynamic that can arise in other shared systems, reinforcing a sense of equity across the community.

“The clarity of this approach has been crucial for building institutional buy-in and researcher trust,” Scharfenberg noted.

While Hellbender’s collaborative access model delivers clear value, the system’s real success is reflected in the breakthroughs it enables. From life sciences to quantum research, Hellbender supports a wide range of technical workflows, including cryo-EM image analysis and GPU-sliced AI modeling. The results speak for themselves, showing what becomes possible when collaboration and access are prioritized and researchers are empowered.

One standout example came during the CASP competition (the Critical Assessment of Structure Prediction), often described as the “Olympics” of protein structure prediction. The MULTICOM team, led by Jianlin “Jack” Cheng from Mizzou Engineering, leveraged Hellbender’s computational power to compete against nearly 100 teams from around the world.

Jianlin “Jack” Cheng (left) and postdoctoral fellow Jian Liu of Mizzou Engineering’s MULTICOM team, whose CASP competition success was powered by Hellbender.

The results were exceptional: first place in estimating the global fold accuracy of protein quaternary structures, third in two additional server prediction categories, and seventh overall among all competitors. 

“It’s a big competition, so it was a dream come true to get our method to become number one,” said Raj Roy, a PhD student on the team.

This kind of success is exactly what Hellbender was designed to support. According to Keeler, researchers often say their work simply “wouldn’t have been possible” with previous infrastructure. “When someone says, ‘I got this paper into Nature because of this resource,’ that’s incredibly fulfilling,” Keeler said.

The benefits extend far beyond large research labs. Through the CIMUSE consortium, Hellbender now supports students and faculty at primarily undergraduate institutions across the state of Missouri. With Open OnDemand, these campuses gain access to advanced computing capabilities that would otherwise be out of reach, bringing high-impact research and learning opportunities to more people across the state.

Thanks to strategic partnerships, thoughtful design, and a strong commitment to shared access, Hellbender’s impact now reaches far beyond any single lab or university.

As Buddy Scharfenberg summed it up: “In short, it has ignited collaboration.”

Missouri’s flagship HPC system strikes a balance between power, flexibility, and accessibility.

• Compute Nodes: 210 (most are AMD with 129 or more cores)
• Total CPU Cores: 24,292
• GPU Nodes: 28 nodes with 100 GPUs, including NIVIDIA A100 and H100
• Memory: Up to 1.3 TB on high-memory nodes
• Storage: 8 PB GPFS and VAST
• Access: Open OnDemand portal with condo model (preemption + shared idle cycles)
• Use Cases: Cryo-EM, protein modeling, genomics, physics, AI training, classroom instruction

Missouri’s Hellbender system demonstrates that transformative HPC is not just about the largest systems. It is also about the most effective approaches to access, governance, and collaboration. SC25 presents an opportunity to explore these ideas, connect with experts, and engage with the global high performance computing community.

Attendees will experience cutting-edge research and real-world applications through technical presentations, papers, workshops, tutorials, posters, and Birds of a Feather (BoF) sessions — all designed to showcase breakthroughs in AI, HPC, and emerging technologies. Join us in St. Louis, November 16–21, 2025, for a week of discovery, inspiration, collaboration, and community.