The State of HPC in Industry

If you spend any time talking with NCSA’s Director of Industry, Brendan McGinty, you’ll quickly learn that he has a great deal of fondness for NCSA. While it’s clear he takes pride in his strong work at the Center, his affection for NCSA goes deeper than that. He was a student at the University of Illinois at Urbana Champaign (UIUC) and worked with PLATO, a first-of-its-kind computerized instruction system developed at UIUC. After graduating from UIUC, McGinty said his first real job was working as a systems programmer for KTH Royal Institute of Technology in Sweden. Eventually, his desire to strike out on his own took hold. He moved into the entrepreneurial phase of his career, starting a number of companies while working as a corporate consultant. All the while, he kept tabs on what was happening at UIUC.

He remembers when NCSA was announced, and his background in programming made him especially intrigued by the idea of publicly funded supercomputing. He has strong memories of when Blue Waters was announced and built, at times speaking about the storied supercomputer as if it was built just yesterday. So when the chance arose to return to UIUC, and in particular, work for NCSA, McGinty took it. “What an honor to go back to my alma mater and serve by bringing companies solutions through HPC,” he said.  “It has been just such a great honor to be back here.”

McGinty brought a unique meld of business and systems knowledge with him when he joined NCSA in 2017. He saw tremendous potential in NCSA, starting with the name of the Center itself. “[Larry Smarr] proposed to the federal government, specifically the National Science Foundation, the concept of a supercomputing network within the United States. I think one of the brilliant things that Larry did was name ours the ‘National Center for Supercomputing Applications.’ That served us well.”

That focus on applications is something McGinty brings up often when he presents at conferences around the world. To McGinty, this is key to how NCSA differentiates itself from other supercomputing centers. He’s not alone in this. NCSA directors have continued to highlight the A in the title. “One of the many things I love about Bill Gropp,” McGinty said, “is that he emphasizes the A in NCSA when he talks about us.”

What is it about the word “applications” that makes NCSA so different? To start with, it emphasizes that NCSA isn’t simply a building that houses supercomputers. “Infrastructure. That may be the least of it,” McGinty explained. “Big machines are everywhere. The EU is currently standing up eight new machines. They’re talking about standing up industry-dedicated machines so that they don’t have the restrictions that other federally funded machines have.” Or, as McGinty succinctly put it at a EuroHPC summit conference this year, “The machines are one thing. But what are you going to do with them?”

NCSA has always been so much more than its machines, and that’s what the word “applications” helps elucidate. A supercomputer represents promise. Applications are the vision of that promise. “The application in our title says, ‘What are you going to do with [the supercomputer]?’ Blue Waters was the biggest machine in the world at the time it was made. It had an incredible shelf life. The reason it had such shelf life was because so many people were able to make good use of that resource to do really cool application work.”

The applications developed on Blue Waters are what continued to keep the machine relevant long after it was created. Most supercomputers aren’t designed to beyond five years. Technology advances make these computers obsolete fairly quickly. New computers run things faster, more efficiently and with less of a draw on the power grid than older models, so it makes sense to move onto newer systems as time goes by, especially as the demands of research increase. It’s a testament to the strength of the developers and Blue Waters’ support team that the machine was in operation for nine years and pushing outstanding results until 2022. 

It’s exactly this type of innovative thinking and expertise that makes NCSA such a good partner for industry needs, according to McGinty. Even from the very beginning, McGinty said NCSA understood the importance of HPC in industry. “1986 is when NCSA opened. And 1986 is when they started thinking about industrial outreach. That is how quickly the National Center for Supercomputing Applications recognized the need.”

But how does a not-for-profit, academically centered supercomputing center appeal to the needs of industry? Industry and academics are often pitted against each other – industry’s focus isn’t the same as academics. “It’s about return on investment for industry,” said McGinty, “which is different than academia.”

In order to continue to attract industry, McGinty knew he needed to show the real value that NCSA could bring to the table. “Companies don’t care about how great we are or the interesting research that we’re doing,” McGinty explains. “They care about the return on investment, how quickly they’re going to get their solutions, how competitive they’re going to be in their sector. They care about that.”

He also explains that the big computers alone are only so much of a draw for them. Big corporations have the money to buy their own machines and, as McGinty mentioned, no restrictions on using them. Academics come to the Center because NCSA offers them something they can’t afford to do on their own. For industry, McGinty works on highlighting all the ways NCSA is more than supercomputers.

“Why come to us?” McGinty said. “Why talk to us? Because we’ve been doing it a long time. Because we have a lot of results to show for that. So, we serve as a bit of a translator.” By focusing on NCSA’s breadth and depth of expertise, McGinty makes a compelling argument for partnering with NCSA over going it alone “because then they don’t have to do it themselves. We can do it for them. We have the expertise.”

“A lot of companies are going through digital transformation. Like with AI. ‘Okay, cool. What is that gonna do for us?’ they might ask. That’s when my pitch really lands. I walk a genomics expert or an AI expert or a modeling expert down the hall to talk with them. And all the details come out from our experts. And suddenly [my industry client] thinks, ‘Okay, now I get it.’ We generate this trust because of the expertise that we have in data and HPC. We align the research that’s happening here with the need there [in Industry] and the magic truly happens.”

Experience counts here. Compute power is everywhere now. Having versatile experiences to know how to maximize compute’s effectiveness is about the people, and there is no place with more experience than NCSA.

 – Brendan McGinty

McGinty also brings up another key benefit to working with NCSA. “We also have talent at the university.” What he means by this is that NCSA’s location and partnership with UIUC is part of the deal. In his EuroHPC summit presentation, he brought up the various training programs running out of NCSA in collaboration with UIUC. Programs like SPIN and REU are part of the attractive package NCSA has to offer. “To have a program dedicated to the HPC curriculum, dedicated to that work? Degrees will come out of that. So forward-thinking, and it will provide for the future like we haven’t seen before.”

“For us at NCSA, it’s more about the people and their expertise than the compute power per se,” said McGinty. “Experience counts here. Compute power is everywhere now. Having versatile experiences to know how to maximize compute’s effectiveness is about the people, and there is no place with more experience than NCSA.”

Being embedded on university grounds also comes with a great deal of potential for industry partnerships. NCSA’s work with academic researchers means they have access to an R1 teaching institution to pull in the best and the brightest – affiliated faculty members and researchers who can lend their knowledge to create solutions. “Let’s say a big clinic approaches us,” McGinty said, “working on drug discovery. We can come in with a genomics expert who can program and who knows HPC – they’re triple threats. That’s the sort of talent that resides within NCSA, and it’s pretty unique.”

McGinty also believes NCSA’s decades of experience and expertise play a large role in the Center’s attractiveness to outside corporations. By positioning NCSA as a place for cyberinfrastructure expertise, the Center becomes the go-to for answers to HPC issues and needs. “In Industry, our team tries to be like a custom consulting shop. ‘What’s your problem? Okay, let’s address that.’ But one of the unexpected outcomes for partners is how we see things from an HPC perspective. For example, we’re not experts on recipes for different octane levels for gasoline. So when I bring in our data experts, they’re coming from a totally fresh perspective. And they come up with some ways to optimize those recipes and gain depth to them through deep learning solutions with some horsepower. And that’s the magical combination. You put a subject expert in a room with an HPC data expert.”

To emphasize the variety of expertise and research domains with issues NCSA could assist with, the Center started with a change to how they approached HPC. “We’ve gone from one big machine to what our director calls a garden approach,” said McGinty.  “So we have, to steal crops term, a garden of resources. You can pick your architecture and customize your solution.” In this sense, NCSA’s changes not only appealed to academics working on their broad spectrum of research in a university setting but also to those working and researching in industry. “The garden approach resonates with industry. And it resonates with what’s happening around the world, as everybody who does research realizes that they also need all of these things: the compute, the expertise on what to do with the compute and the expertise on how to best use the data.”

Of all the things McGinty does to continue NCSA’s productive relationship with industry, there’s one part he makes sure never to skip out on. “Don’t forget to finish the job and track how effective it’s been for companies because when you show that they spend X dollars and you save them so many more, they’ll be back for sure.”

McGinty’s work with industry obviously benefits the companies he works with, but some may be wondering what the benefit to NCSA is. In short, NCSA provides industry with solutions, and industry gives NCSA researchers and programmers valuable experience working with new and interesting challenges. Each party comes away learning something useful. Much of what McGinty’s team works on deals with issues that companies are facing today, and solving these issues not only gives NCSA researchers and programmers new ideas and inspiration on how to tackle some of their other projects but also allows them the opportunity to see the impact of their work much more immediately.

“Industry collaborations have been important to our Center for several reasons,” said McGinty. “Unique, applied datasets with real-world challenges and opportunities are among them. Our program was formed many years ago to address American competitiveness. It’s still about that, but now on a global scale.”

This same spirit of knowledge sharing is why McGinty also spearheaded the International Association of Supercomputing Centers (IASC). Formed with the goal of bringing the nearly 100 publicly funded supercomputing centers together in the hopes that the shared knowledge will help individual member institutions resolve issues and optimize their services, IASC is still in the early stages of fruition. The founding members include the UK’s Hartree Centre, Lawrence Livermore National Laboratory (LLNL), NCSA and the Leibniz Supercomputing Center. The group has just begun sending out formal invitations to join and has been quietly planning, waiting until they had something to offer before going public. But McGinty has high hopes for the nascent group.

“We want to share the knowledge of our common issues and resolve them,” he said. “[HPC professionals] work in too many silos. There are not enough resources to do these grand challenges and address them as a planet. And they’re huge things, be it climate, energy, sustainability, or speed of care to the bedside in healthcare. The Swiss are just as dedicated to these issues as the Japanese, as we are – as anybody is. And we have a responsibility to prepare the next generation of HPC experts, and that means looking worldwide. We need more diversity and more youth. And this is one of the things that we’re going to talk about at the international level by bringing these groups together.”