From cloud sprawl to strategic success: Cornell University’s cloud service transformation

Within a decade of moving to the cloud, Cornell University in Ithaca, New York, was managing 260 accounts—160 of them on Amazon Web Services (AWS)—at an annual spend of $4 million. As part of what they called a “cloudification initiative,” there had been a major push in the early days to capitalize on opportunities to move applications into the cloud and support innovation across the highly decentralized institution. But the rapid adoption had led to cloud sprawl, without clear guidelines for ongoing maintenance and support.

The root of the problem lay in the very terminology the university used. “An initiative is not a service,” said Chris Manly, Assistant Director of Cloud and Infrastructure Services at Cornell until October 2025. “A service has a definition of what you do and what you don’t do. And we didn’t really have that. We didn’t have clear boundaries and scope.”

Chris realized that to establish a plan for Cornell’s next decade of cloud use, he needed a structured approach to evaluating their cloud estate and capturing customer feedback. In late 2024, Cornell began a comprehensive service exercise to align its cloud operations with institutional needs. This post outlines the approach Cornell took to improve its cloud operations by listening to customers, identifying key gaps, and setting clear priorities to build a sustainable service plan that will carry it into the future.

From championing adoption to providing direct support

Cornell was among the first higher education institutions to see the potential of the cloud in supporting enterprise computing, academic research, and teaching and learning. In 2015, they signed a direct agreement with AWS to facilitate its growth in the cloud. Also that year, Cornell’s then-assistant director of cloudification services and now Deputy CIO, Sarah Christen, founded the Cloud Forum to bring together higher education IT professionals to explore the opportunities, challenges, and best practices of using the cloud. She was recognized as the 2024 Cloud Superhero by Internet2 for her leadership role in the higher education cloud community.

Although Cornell was leading the way in cloud enablement, the university’s highly decentralized structure made it difficult to implement and enforce a common set of principles around ownership and maintenance of cloud-based services. In addition to a central IT office, many of Cornell’s administrative units and colleges have their own IT function, with some being light users and others, such as the Cornell Lab of Ornithology and the College of Veterinary Medicine, being heavy users of the cloud.

Over time, as IT staff responsible for cloud applications in their individual units left the university, central IT was called in to provide direct support. “We were going beyond the enablement mode and the brokering into acting as a managed service provider and actually doing hands-on support,” explained Manly. It was a service model that quickly became unsustainable.

Engaging end users to build a new service design framework

Recognizing the need for a reset, Manly worked with AWS to undertake a comprehensive service design exercise. He wanted to understand what was working, where there were gaps, and what changes were needed. Manly started by talking to central IT leadership, then to the engineers on the central cloud team. Then, importantly, he met with a group of campus stakeholders and said, “Here’s what we’re seeing in central IT. What do you see? What do you need?”

This process uncovered some key strengths of Cornell’s existing cloud management. In general, there was consensus that the enablement approach works. The central IT team is responsible for setting up accounts and establishing guardrails, then handing over the keys to users who have open access moving forward. Another strength is the cloud team’s ability to solve problems, streamline processes, and automate workflows. One of the engineers joked to Manly that “every day we’re trying to code ourselves out of a job.”

Finally, leaders and stakeholders agreed that they were stewarding the university’s resources effectively thanks to the partnership with AWS. The team maintains 90% Reserved Instance coverage for AWS and 70 percent for Amazon Relational Database Service (Amazon RDS) through automated savings plan management with Spot ECO. The savings then get passed back to individual departments.

Clamoring for improved communication with stakeholders

The feedback sessions with stakeholders also identified some key areas where Cornell needed to improve its cloud service plan. Perhaps the biggest surprise to Manly was the desire for more communication and community building among cloud users. In the early days, he explained, the central IT cloud team had open office hours 1 day a week and an actual room where anyone on campus could come for technical support or troubleshooting in real time.

“We didn’t make the transition from an early-stage startup culture to a more mature model,” Manly said. “And then, with the disruption of the pandemic, the collaborative environment we had put together just kind of fell apart.”

Additionally, regarding communication, stakeholders expressed an interest in enhanced knowledge sharing. Although the cloud team answered questions well, they weren’t as proactive about sharing lessons learned across campus as people would have liked.

On the technical side, stakeholders identified the need for more support in some key areas, namely: ensuring compliance with data security and privacy requirements; optimizing their specific budgets for efficiency (FinOps); and carrying out routine monitoring, log management, and other functions that had existed in data centers but weren’t replicated in the cloud. They also mentioned struggles with Amazon Elastic Compute Cloud (EC2) patching and AWS Identity and Access Management (IAM) role management.

Setting the stage for a cloud center of excellence

The consensus around the central IT cloud team’s strengths clarified its role and purpose. “What we do is core cloud enablement and managing the relationships with the vendors,” said Manly. “If someone comes to us with a problem, we’ll provide some support, but essentially we’re teaching someone to fish.”

To address the gaps in support identified by stakeholders, the central team is rebuilding the campus cloud community with stakeholder-led governance. The goal is to create a cloud center of excellence, but Manly acknowledged that it is a way down the road. The central team is also making the case for creating a dedicated FinOps staff position to save money in the long term. And finally, the team is working with IT security staff to develop a compliance strategy, starting with documenting the current landscape and identifying the greatest needs.

Having seen the benefits of a service design exercise, Cornell has now committed to adopting a continuous iteration mode moving forward. This will help the university stay up to date with new enhancements to the cloud and make sure that its cloud usage stays in line with institutional needs.

Learn how AWS helps institutions build, deploy, and scale cloud solutions that address public sector needs. Contact us today at AWS Public Sector.

Read related stories on the AWS Public Sector Blog:

• Vanderbilt University launches Cloud Innovation Lab in collaboration with AWS
• Macquarie University accelerates cloud transformation with AWS
• Educating for a greener future: Realizing sustainability gains with AWS Cloud migration
• Seven lessons learned: Driving AWS Cloud adoption in research at Saint Louis University