How Dartmouth College Leveraged AI To Modernize Its Network in One Weekend

Dartmouth College, one of the smallest but most prestigious Ivy League schools, recently embarked on a significant network upgrade to provide students and faculty with faster, more reliable campus-wide connectivity. The university, which supports approximately 25,000 simultaneous users, partnered with Juniper Networks to deploy Mist AI, an AI-powered platform that manages wireless networks using real-time insights, automation, and troubleshooting.

A crucial part of the upgrade was deploying 2,000 Mist access points (APs). The university only had two days to install the APs, as students returned to in-person learning in the fall of 2020 following the COVID-19 pandemic. Dartmouth received the newly released Mist AP12 on August 19, but cleaning crews were slated to sanitize all dorm rooms starting August 22. Once the rooms were cleaned and sealed, they could only be opened by students in September. Hence, Dartmouth faced a significant challenge in completing this large-scale installation on a tight schedule.

“At the time, we had already started to deploy Mist, and we were coming up with plans to roll it out to the rest of the campus. We realized we’ve got a tight window to make this happen as students were planning to return. We worked closely with the Juniper team to install small form factor wall plate APs—click onto the dorm room wall, easy to install, and provide Wi-Fi at a great cost,” said Bryan Ward, Dartmouth’s lead network engineer, in an interview with ZK Research.

APs Deployed and Configured Rapidly Using AI

The networking team, consisting of five engineers, configured each AP to ensure consistent performance throughout the campus. According to Ward, the setup took only a few minutes, which is a game-changer. Volunteers from other Dartmouth departments, including non-technical staff, also helped with the installation. From my experience as a network engineer, the typical time to set up an AP using manual configuration methods is 30-45 minutes. Without AI and automation capabilities, deploying 2,000 APs would have been impossible.

Integrating the new APs into Dartmouth’s existing network infrastructure was no small feat. Mist’s cloud-based architecture and AI-driven insights made this easier. The platform provided real-time visibility into network performance so the networking team could monitor progress and adjust when needed to avoid service interruptions. The team used Mist’s insights to identify gaps in coverage, especially in high-demand areas. These included dorms and the newly opened engineering and computer science facility, which required approximately 200 additional APs.

Dartmouth’s network has grown rapidly in recent years due to increased users, new construction, and the demands of research facilities. For example, the university recently activated two new 100-gigabit internet circuits to handle the increasing load.

“We were bumping up against the limits of our existing service. The uses are growing more than anybody could have predicted. With the increased use and load on the network, we’re just trying to stay ahead before it becomes a problem,” said Ward.

AI Benefits Extend Beyond Deployment

Following the deployment, Mist’s AI capabilities continue to provide value. Marvis, the platform’s virtual assistant, continuously scans the network, diagnosing problems and recommending solutions in real time. Marvis has become an essential tool for the networking team. Tasks, such as analyzing logs, identifying anomalies, and diagnosing root causes, that previously took hours, can now be completed in seconds.

Ward, who was initially skeptical about AI's role in network management, now sees it as an invaluable tool. Marvis has streamlined many aspects of network troubleshooting and maintenance, allowing Dartmouth’s small IT department to focus on more critical tasks rather than spending time on manual troubleshooting.

“I view Marvis as a powerful, fast assistant, not a job replacement. It has only helped me in my job since I’ve been using it. My help desk staff can also log into Mist and ask Marvis the same questions I can. Marvis will provide that information in a human-readable format, not just ones and zeros,” said Ward.

Looking ahead, Ward envisions a future where end users can interact directly with Marvis to resolve network issues, such as a poor Wi-Fi connection, without the help desk getting involved. Junior engineers and help desk staff could also leverage Marvis to handle issues without escalating to senior engineers. This would create a more efficient, user-friendly experience for students and faculty while easing IT’s immense workload.

Dartmouth's success with using AI is a good lesson for IT pros. Many engineers I have interviewed are nervous about AI taking their jobs. The reality is that AI is a tool, and no engineer will lose their job to AI. What’s more likely is they will lose their job to another engineer who has embraced AI and can work at a pace that enables the business to meet its goals.

Zeus Kerravala is the founder and principal analyst with ZK Research.

Read his other Network Computing articles here.