“We’ve been pushing a lot of our researchers to use our on-premises research infrastructure,” says Zachary Rossmiller, the university’s associate vice president and CIO. “But now we’re getting to the point where some of it’s at capacity, so we are looking at ways we can expand and burst into the cloud.”
AI workloads complicate the picture for R1 IT leaders. Hard-and-fast rules for cloud-versus-on-premises workloads don’t cut it anymore.
“IT teams increasingly treat cloud and on-premises as a unified research platform with common tooling and schedulers to route entire workloads to the environment best suited for the job,” says Matt Jubelirer, general manager of product marketing at Microsoft Education. Software helps pick the optimum workload location based on performance requirements, cost models and data locations. Hybrid schedulers and orchestration tools help researchers direct workflows to on-premises systems or to cloud resources, depending on workload needs, capacity and timing, he adds.
Robust bandwidth is becoming mandatory.
“We’re prioritizing instant and sustained connectivity over simple speed,” says Douglas Little, Georgetown University’s CIO. Georgetown recently upgraded to Wi-Fi 7 to ensure students and researchers have wireless connectivity for AI and other bandwidth-hungry applications.
DISCOVER: Data center as a service bridges the gap between on-prem and cloud infrastructure.
Georgetown’s medical researchers, meanwhile, need low-latency, high-capacity networks to run virtual and augmented reality (VR/AR) simulations, Little says.
R1 universities often lean heavily on research computing facilitators who help them weigh these decisions. “One good facilitator can steer dozens of researchers, educators and their research teams to the right mix of resources, and that saves money that otherwise would go in the wrong direction,” says Dana Brunson, executive director for research engagement at Internet2, which provides a suite of technology and collaborative resources to the higher education sector.
Internet of Things programs and remote sensing projects are moving AI workloads closer to data sources, Brunson adds. Researchers and first responders, for instance, can use AI to study imagery of remote mountains to detect current wildfires and anticipate where they might flare in the future.
Where R1 Infrastructure Is Going in the Future
High-performance computing is rapidly evolving amid surging demand for newer hardware with advanced graphics processing units. Researchers want on-demand access to AI resources, whether in the cloud or on-premises.
LEARN MORE: Modern storage environments protect sensitive data.
“The infrastructure conversation is shifting from ‘how big should our cluster be?’ to ‘how do we offer the right level of compute, platform or service for very different research needs without overbuilding on-premises systems?,’’’ Jubelirer says.
Expect demand for IT infrastructure to keep surging with the rise in AI workloads, Rossmiller says. Data sovereignty and cloud costs will be increasingly important in states with rural communities that need access to powerful computing. “Knowing where their data lives and sits is crucial to them,” he says. Rising cloud costs create incentives to keep workloads on-premises and in-state.
The computing conversation will go far beyond software, hardware and the challenges of well-resourced R1 schools, Brunson says. “The vast majority of the students in this country are not at R1s,” she says. Students at R2s and at teaching-focused, minority-serving and primarily undergraduate campuses can’t be overlooked. “We want to make sure they have the opportunity to become the scientists, the workforce for tomorrow.”
Actionable Advice for Campus Leaders
R1 IT leaders should view their network as a mission-critical foundation, Little says. He encourages modernizers to start where the demand is heaviest. “Focus your initial rollouts on areas where users naturally congregate, like residence halls and large lecture halls, to ensure the network can handle the modern ‘stadium effect’ of multiple high-bandwidth devices.”
UP NEXT: Infrastructure modernization helps support future technologies.
Be realistic about the cultural obstacles that come with moving to new technology platforms, Rossmiller says. “If you’re going to go all in on cloud and Software as a Service, then you need to have a really strong change management program because a campus is really hard to change and shift and be agile,” he adds. Rossmiller also encourages building a strong vendor-management team.
Most of all, work together. “The biggest opportunity is in the people,” Brunson says. Campus research computing facilitators can be force multipliers who help IT teams adapt to changing resource demands, she adds. Confer with leaders at national resources like the National Science Foundation’s National Artificial Intelligence Research Resource (NAIRR) and Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) and at local organizations like Georgetown’s Center for New Designs in Learning and Scholarship.
“Technology’s changing so fast that no single person can possibly keep up,” Brunson says. “Every campus benefits from being part of a larger conversation.”
