Sumadi Scales from 2,000 to 50,000 Concurrent Users after Migrating to AWS

Educational technology company Sumadi, which specializes in automated test proctoring solutions, needed a way to differentiate its offerings by delivering near-real-time supervision so that customers wouldn’t have to wait hours before receiving the results of an exam session. Sumadi also wanted to be able to support an increasing number of concurrent users, perform near-real-time processing, and queue exam session results.

To achieve near-real-time supervision simply and cost effectively, the company decided to migrate to Amazon Web Services (AWS). Now, Sumadi can deliver its services in near real time and can scale to meet its customers’ growing need for online test proctoring solutions that fit how education increasingly takes place today: remotely, flexibly, and on demand.

Operating since 2019, Sumadi is based in the Netherlands and runs its corporate operations from Honduras. The company primarily works with educational institutions and other customers throughout Central and South America and Europe to make online test taking more secure. The company’s automated proctoring solutions—powered by artificial intelligence—use facial and typing-pattern recognition, authentication, object detection, screenshots, audio recording, and proprietary browser tracking to secure end-user devices and monitor online exam sessions. “Online learning is increasingly becoming a larger part of the total share of educational and instructional hours of institutions around the world,” says Francisco Pérez, director of product development at Sumadi. “What really drove us at the beginning was to provide a cost-effective solution with the necessary scaling and near-real-time capabilities for institutions wherever they are.”

Sumadi, which was born in the cloud, used a multicloud infrastructure that included a third-party vendor but was not satisfied with the level of performance and customer service that it received. Because Sumadi already used several AWS services, the company decided to migrate its database to AWS and was able to complete the process in just 3 weeks. “We’ve experienced a world of difference by using AWS,” says Pérez. “Our account is managed not according to the size of our business but to the problems we are trying to solve. The service we receive from AWS has been game changing.”  

Sumadi finished migrating to AWS in August 2020 after testing various cloud services and deciding to use AWS. During early testing, Sumadi started using Amazon Rekognition, a service for automating image and video analysis using machine learning. The company now uses Amazon Rekognition models for both face and object detection as part of its proctoring solution. For example, Sumadi can use Amazon Rekognition to detect if a student is taking an exam on another’s behalf or if test takers have cell phones, calculators, or other objects that are prohibited during exams. “From the start, Amazon Rekognition provided the accuracy that we needed,” says Julio Elgueta, enterprise architect at Sumadi.

The company started off small, using regular servers, and tested its first exam on just 1,000 users. However, Sumadi needed to be able to scale to serve many thousands of users concurrently. To do so, the company opted to switch to serverless technology using AWS Lambda, a serverless, event-driven compute service that lets developers run code for virtually any type of application or backend service without provisioning or managing servers. Sumadi uses AWS Lambda functions to process messages in queues within Amazon Simple Queue Service (Amazon SQS), a fully managed message queuing service. The company can now ingest millions of images in a single day.

Sumadi stores this vast quantity of data in two databases. For storing image metadata, the company uses Amazon DynamoDB, a fully managed, serverless, key-value NoSQL database designed to run high-performance applications at virtually any scale. For standardized data used for exams, courses, students, and colleges, Sumadi uses Amazon DocumentDB (with MongoDB compatibility), a scalable, highly durable, and fully managed database service for operating mission-critical MongoDB workloads. MongoDB is an open-source, NoSQL document database. Using this combination of two databases, Sumadi can store millions of records, process the records in near real time, and present reports to supervisors or teachers reviewing the exams.

Since migrating to AWS, Sumadi has been able to cost-effectively scale its proctoring services to support 50,000 concurrent users, each generating one record every 30 seconds—an increase of 25 times compared to the 2,000 concurrent users Sumadi was able to handle in 2019. By being able to serve educational institutions and students with fully remote exam proctoring, Sumadi offers the flexibility needed to expand educational access beyond physical locations. Many institutions needed to shift to online learning due to the COVID-19 pandemic, and Sumadi provides a means of maintaining integrity during online remote assessments. Administering exams remotely is also a more cost-effective solution—especially for organizations such as the Colombian Institute for Educational Evaluation, which proctors high school exit exams for as many as 300,000 students at one time. “The logistics of testing that many people in one physical location are complex and expensive,” says Pérez. “Using AWS, we’re able to streamline the whole process, make it cost effective, and monitor hundreds of thousands of users concurrently in multiple countries, while providing results in near real time.”

Now that it has migrated to AWS, Sumadi has gained the ability to offer remote exam proctoring in near real time and at scale. The company is eager to continue acting as an early adopter of AWS services—especially in the machine learning space—so that it can enhance the user experience and deliver premium features to its customers. For example, using WebSocket technology support provided by Amazon API Gateway—which lets developers create, maintain, and secure APIs—along with AWS Lambda functions, Sumadi has developed a chat application that test takers can use to communicate with supervisors. The company intends to integrate this feature with a more comprehensive exam solution that it plans to build on AWS.

Sumadi also intends to explore ways to use video—along with the continual snapshots it currently offers—during exams, which will deliver a higher level of inferencing and provide customers with a range of choices to suit their needs and budgets. “Using AWS empowers us to bring the latest advancements to our customers,” says Pérez.