By Katie Elyce Jones, PillarQ
In December 2022, the Ohio Supercomputer Center (OSC) located in Columbus celebrated its 35th anniversary “at the forefront of technology” as an academic and industrial high-performance computing (HPC) and computational science resource. The center serves more than 8,000 clients from hundreds of educational institutions, nonprofits, government agencies, and companies.
In an anniversary press release, OSC cited past research successes, such as sequencing DNA data to help grow more disease-resistant crops, and future endeavors, including workforce development in anticipation of a nearby Intel facility that will be “one of the world’s largest microchip manufacturing complexes.”
Today, the forefront of technology also includes emerging fields like artificial intelligence (AI). OSC Director of Research Software Applications Karen Tomko said the center now provides a variety of computing and storage resources for AI.
New eras in HPC
Many government and academic supercomputing centers can relate with the anniversary milestone.
The 1960s and ‘70s saw the rise of U.S. supercomputing, with commercial computers built namely by IBM and Cray. Federal agencies and big research institutions commissioned these state-of-the-art systems, seeding future supercomputing centers. In the mid-80s, the National Science Foundation (NSF) established five core supercomputing centers across the country.
Around the same time in 1987, the OSC program was established by the State of Ohio as a statewide resource, enabling the center to engage industry users in a way some university-based centers could not. Across the country, the HPC community grew. The Association for Computing Machinery and the IEEE Computer Society hosted the first SC (Supercomputing) conference in 1988. Five years later, the Top500 project released its first ranking of top supercomputer performance.
However, much of this HPC expansion in the final quarter of the 20th century ran parallel to another computing era—a series of “AI winters” as funding cuts and other setbacks intermittently put AI technology on ice. But AI research has heated up in the 21st century as an explosion of big data and processing power drive new research. Now, HPC centers like OSC are increasingly incorporating modern AI advances into their services.
OSC and a future of AI advances
OSC recently launched its Ascend system, a Dell-built HPC cluster accelerated for AI with NVIDIA A100 GPUs.
“Ascend was developed to respond to the growing number of clients seeking to use AI in their work,” Tomko said. “One of our clients who used Ascend during our early testing period reported being able to use a massive neural network model for natural language research for the first time, thanks to the power of Ascend’s GPUs.”
Ascend. Image credit: Ohio Supercomputer Center
The center also doubled the amount of data it can store and tripled the speed of its storage system in the last year to accommodate data demands.
In addition to compute and data resources for AI, OSC is a core collaborator in applying AI to key industries and research domains. As part of the NSF-funded AI Institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) led by The Ohio State University (OSU), OSC and 13 other partners “will build and prove a next-generational AI system around the application domains of smart foodsheds, digital agriculture, and animal ecology,” Tomko said. OSC is one of three supercomputing centers providing ICICLE with HPC and large-scale data storage.
When asked what OSC sees as the role of AI in supercomputing and user research in the next 5 to 10 years, Tomko referenced chip and system designs for AI workloads, data center cooling and power management that employ AI techniques, workload analysis and management, software development tools (such as code generation), and the growing use of conversational interfaces (such as for user support).
Beyond hardware and data center management, Tomko said “engineering and science domains have been quick to leverage AI” with applications such as protein structure, autonomous vehicles, and the analysis of unstructured texts.
“Ubiquitous data collection means there is data available that is relevant to every area of study from agriculture to public policy,” Tomko said. “Supercomputing use in all fields is growing. For example, we had a research software engineering project with a researcher who does analysis of soil properties and is developing AI models based on spectral data of soil samples.”
Other AI research projects using OSC resources include natural language processing research, led by Huan Sun of OSU, to examine problems from foodsheds to health care, as well as drug discovery research led by Xia Ning of OSU and more.
Perhaps the next 35 years will not only see progress in AI but other emerging technologies impact HPC centers like OSC.
HPC centers aren’t just pursuing AI missions but missions to advance the emerging field of quantum information science (QIS) as well. Read more from PillarQ about HPC and QIS integration at the Oak Ridge Leadership Computing Facility and the National Energy Research Scientific Computing Center.