Schools are also embracing small learning models that move masses of data closer to computers. “These SLMs are purpose-built,” Lee says. “They’re perfect for domain-specific tasks without all of the massive overhead of cloud-based large language models.”
A data analytics specialist, Lee also focuses on building interactive, AI-empowered learning environments. His post at Miami Dade places him at one of the great U.S. higher education success stories: a school that grew from a junior college in the 1960s into a four-year institution that now boasts eight campuses and more than 100,000 students.
Lee says on-premises SLMs also reduce latency that degrades application performance in cloud-hosted apps. He describes apps using “emotionally intelligent AI” that can adapt to student learning styles in real time. These kinds of apps require ultralow latency that remote, cloud-based infrastructure has a hard time delivering, he added.
One worry spanning on-premises and cloud is shadow AI, where people spin up AI apps without oversight from central IT. Lee cautions that shadow AI can happen anywhere — from a smartphone or a laptop to an on-premises workstation — with zero visibility for IT or cybersecurity authorities. “These small models have access to people's data locally — their email, their calendar, their file system, etc.,” he says. “It's a very big problem, and it's growing.”
DISCOVER: Managing AI and cloud complexity doesn’t have to compromise security.
How Universities Manage Cloud Costs and Complexities
Historically, many hyperscale cloud providers offered credits to higher education institutions that made cloud-hosted workloads cost-effective, Lee recalls. Recently, those credits have gone away. “When the cost started to rise, we were left in a really uncomfortable position,” he says.
Moreover, cloud migrations are inherently massive projects. “Education leaders have to evaluate the actual time it takes to get workloads ready to migrate to a hyperscaler,” says Ariel Obando, systems engineer for state, local and education at Nutanix, whose technologies help institutions manage hybrid environments. “They also have to consider replatforming, which varies wildly depending on the specific workloads,” he adds.
Obando consults with colleges and universities in the U.S. Northeast. “As one of my customers puts it, ‘cloud where appropriate’ is what's important,” he says.
CDW’s technology services team saw this firsthand in a recent, cloud-first project modernizing the infrastructure, strategy and operations of a large public university. “It's not an immediate, ‘Hey, we did this in six weeks,’” says Michael Durand, director of higher education sales at CDW. Cloud transitions can last weeks, months or years. “It’s a tremendous cultural change for the IT department,” Durand says.
LEARN MORE: Optimize storage and compute for AI readiness.
Managing Cloud on a Small Campus
A niche college in St. Louis shows how smaller schools weigh the pros and cons of cloud infrastructure. “We focus on resilience in making these decisions,” says Zach Lewis, CIO and CISO at the University of Health Sciences and Pharmacy, which recently was acquired by Missouri’s Washington University.
Applications requiring near-constant uptime have been moved to cloud, Lewis says: “For workloads that are small or need to interact with devices on-premises — think building automation, security or camera systems, alert systems — we keep those here to offset the cost of running in the cloud and ingress/egress charges.”
Lewis wrote a book about a cyberattack at his school that underscores the risks higher education IT leaders deal with. “If we had been fully on-premises, things would have been a lot worse,” he recalls. “Our cloud environment remained unharmed during the incident.” The attack generated more interest in moving workloads to cloud or Software as a Service so they could be segmented to operate independently, he added.
Balancing Cloud and On-Premises Infrastructure
These tips can help IT leaders find the right mix for their hybrid environments:
- Partner with vendors who can upskill your workforce to master the nuances of orchestrating hybrid, multicloud ecosystems.
- When tight budgets bring cost-cutting requests, make sure financial authorities understand the consequences of turning off vital, popular services.
- Look to cloud for accessibility, scale and business continuity, particularly for student-centered workloads that work across multiple channels and devices. Consider on-premises for performance, low latency and domain-specific AI.
- Build cross-functional centers of excellence so IT teams, academic leadership and financial authorities operate from the same playbook.
“As a leader in this space, you have to be comfortable managing a blended environment that's going to optimize cost, performance and security,” Lee says. “Ultimately, whether a workload is in the cloud or on-premises, the guiding question has to be, how does this improve the students’ experience and prepare them for the workforce?”
