Migrating to M7i-flex Instances to Reduce Costs by 20% with Imagine Learning

Imagine Learning is a leading provider of high-quality K–12 (from kindergarten to twelfth grade) curriculum materials, serving 18 million students in more than half the school districts in the United States. Its products include core instruction in essential subjects like math and reading, as well as supplemental programs for literacy support, remediation, and expanded learning.

To support its operations, the company is always looking for ways to reduce costs and optimize its resources. Imagine Learning conducts quarterly infrastructure assessments to make sure that it uses the most appropriate and cost-effective technology for each application. During one of its analyses, the company realized that some of its workloads do not use compute evenly throughout the day. There are periodic traffic spikes each morning as the school day begins across different time zones. “It creates a scaling storm where our compute usage increases rapidly,” says Christopher Moudy, vice president of engineering at Imagine Learning. “Then the peak stays stable while everyone’s in school, and then it declines until the evening. You see a nice rollercoaster curve every day.” This traffic pattern made Imagine Learning’s applications the perfect candidates for using M7i-flex Instances, which the company learned about through its technical account manager. “We were keeping our CPU usage fairly low on purpose to facilitate scaling,” says Moudy. “So, having Flex instances became very interesting to us.”

Imagine Learning decided to evaluate the cost benefits of upgrading its M6i Instances to M7i-flex Instances. The company identified its Imagine Edgenuity application, which has the highest number of daily users, as a good candidate for its trial of the Flex instances. Prior to migration, Imagine Learning tested the M7i-flex instances in multiple environments. “We did our quality assurance environments first,” says Moudy. “We saw how well that worked, and that’s when we got excited.” The company was able to migrate to M7i-flex Instances within a few weeks.

M7i-flex instances are at the hub of every request that comes into Imagine Edgenuity, so the cumulative performance and efficiency boost was substantial. Imagine Learning was able to reduce both the application’s total instance count and its average daily costs by 20 percent. Another unexpected benefit of the upgrade was a 40 percent reduction in the application launch time.

This increased performance and the accompanying reduction in launch time made it possible for Imagine Learning to fine-tune the application’s scaling parameters. “Before, we would target 30 or 35 percent CPU usage, for example, to be confident that we would be able to scale fast enough,” says Moudy. “Now, we’re better able to fine-tune our auto scaling groups to run at a higher CPU percentage, upwards of 75 percent. Being able to smoothly scale and take advantage of that type of horizontal scaling is very beneficial.”

After seeing the cost and performance benefits of the Flex instances, Imagine Learning has upgraded its M6i instances to M7i-flex Instances across all its applications. For some compute-heavy workloads, Imagine Learning has begun using Amazon EC2 C7i-flex Instances, which are compute-optimized instances powered by 4th Generation Intel Xeon Scalable processors.

 

As Imagine Learning continues to innovate and build new education solutions, it appreciates the support that AWS provides. “It helps when you have a dedicated account manager, technical account manager, and solutions architect who really know our products,” says Moudy. “We form relationships over time. It’s the collaboration between Imagine Learning and AWS that makes it even more powerful on top of the great services that are being delivered.”

 

Using M7i-flex instances has helped Imagine Learning continue to provide a high level of service to its users, improving application performance while reducing costs. “AWS provides services that help us deliver really good user experiences by not having to worry about the infrastructure,” says Moudy. “We don’t want to think about a server going down. We want to build applications for students, teachers, and parents. That’s what we want to focus on.”