PowerSchool Adopts a Serverless Strategy to Deliver Personalized Learning

Assessment grading and performance analytics are crucial parts of modern K–12 education. Through its Performance Matters assessment and analytics solution, K–12 education technology company PowerSchool delivers fast scanning and grading services for digital and paper-based assessments. But as the company grew from 2018 to 2020, it found its on-premises data center overwhelmed by predictable bursts of activity as thousands of teachers scanned tests on their lunch breaks and during exam weeks. The company needed to find a solution that would address the increasing costs of high server capacity and enable it to scale its data center quickly enough to meet demand.

Facing long processing backups and expensive server management, PowerSchool knew it needed a more powerful and elastic solution. The company migrated its application stack to Amazon Web Services (AWS) and adopted a serverless strategy to scale effortlessly and deliver near-real-time performance to customers at peak demand while optimizing costs.

From 2018 to 2020, the company’s Performance Matters product grew from serving 150 to 800 large K–12 clients. That kind of rapid growth further strained an already overburdened data center. PowerSchool couldn’t maintain a cost-effective balance between long periods of largely idle server operation and short spikes of high activity. The company was manually scaling up and down around peak usage. “When we knew testing season was coming up, we’d have operations engineers manually scaling up the servers, configuring them, and testing them to make sure they properly processed scans,” says Jason Ergle, PowerSchool’s director of engineering.

This was error prone and costly, and customers experienced processing backups as long as 12 hours during peak activity. PowerSchool prided itself on its ability to scan and grade quickly, but it wasn’t delivering the fast experience on which its users—teachers facing their own deadlines—depended.

In April 2020, PowerSchool migrated to a serverless architecture using AWS Lambda, a service that lets users run code without managing servers and scales automatically to meet demand. “The first slice that we wanted to do—our most valuable product—was getting the scan engine swapped out,” says Michelle Malaski, lead software engineer at PowerSchool.

The scan engine, which reads, analyzes, and grades the tests, is the heart of the stack. After 6 months, PowerSchool had migrated its scan engine and nearly all its data processing functionality to AWS. This instantly removed the barriers to growth that its traditional data center had created. However, the company’s scanning pipeline and other systems had been designed for the constraints of its legacy data center. To realize even more significant cost reductions and take full advantage of the power and flexibility of AWS Lambda, PowerSchool rebuilt its processing pipeline for the cloud.

Using a serverless architecture built on AWS Lambda enables PowerSchool’s processing capacity to scale continuously with a workload’s size, providing the flexibility needed to meet customers’ bursts of demand. This elasticity brings massive savings while enabling fast performance. Running its data center had cost the company $15,000 per month. During peak workloads, the former system could take up to 30 seconds to process each scan. Using AWS Lambda, PowerSchool has decreased that rate to 12 seconds per scan while under load. It has also reduced its monthly cost to just $200. “AWS has enabled us to save 97.6 percent on our operations expenses,” says Ergle. “We have essentially unlimited capacity to maintain near-real-time performance.”

PowerSchool’s scan engine is responsible for unpacking the scanned files, using optical character recognition to read them, and pushing the processed data to its backend database for storage and analytics. Before AWS, the system maxed out at 150,000 scans per hour. Now PowerSchool has increased transactional throughput to 600,000 scans per hour. The company has further optimized its pipeline using AWS Step Functions, a serverless function orchestrator for sequencing AWS Lambda functions. AWS Step Functions makes PowerSchool’s application stack more transparent and pinpoints errors as they occur.

Using Amazon Aurora, a relational database built for the cloud, also drastically reduced the cost of storage for historical grading data. “The Availability Zones of Aurora enable us to run our software seamlessly across multiple data centers,” says Ergle. It also continuously backs up to Amazon Simple Storage Service (Amazon S3), an object storage service that offers industry-leading scalability, data availability, security, and performance. Using Amazon S3, PowerSchool reduced its monthly storage expenses from $6,000 to $240, enabling it to store data for as long as customer contracts require.

PowerSchool’s support of both digital and paper-based assessments makes the company’s services available to many school districts that lack the infrastructure for digital testing. Some school districts can’t afford the devices and technical personnel necessary to roll out online assessments, and others don’t want to put students who lack access to technology at a socioeconomic disadvantage. Moreover, many schools choose to give paper-based exams to prepare their students for the experience of taking similar state and national standardized tests. Using the power and elasticity of AWS Lambda, PowerSchool now has the capacity to provide high-performance scanning capabilities to more schools that depend on paper-based tests. Through better data and analytics, the company’s Performance Matters solution enables more personalized learning and more insightful reporting for students and schools around the world.

The next pieces of PowerSchool’s modernization using a serverless infrastructure are set to be completed in the first half of 2021. It plans to continue expanding its use of Aurora to migrate more of its application stack to serverless architecture. “We’re excited about using this design pattern as a basis for more continuous integration and development,” says Malaski. “It’ll help us move a lot faster and do more with our system.”

PowerSchool also has several projects lined up to further improve availability, scalability, and cost efficiency. “It’s possible, when using serverless technologies, to have your cake and eat it too,” says Ergle. “We can give customers a more positive experience, better availability, and higher performance while also reducing operations and maintenance costs.”