Industry: Education Technology (EdTech)
Goal: Replace Firebase with a globally scalable AWS backend to improve real-time collaboration and support growth
OpsGuru Services: Cloud Architecture Redesign, Re-platforming, Performance Engineering, Load Testing & Validation, DevOps Enablement
AWS & Tooling: Amazon ECS (Fargate), Java Spring Boot API, VueJS Frontend, Redis via Amazon ElastiCache, MongoDB Atlas (Sharded), Amazon CloudFront + ALB, Grafana k6
50,000+ writes/sec performance under test
Low-latency, globally distributed real-time collaboration
Simplified infrastructure and observability
Ready-to-integrate AWS architecture for global rollout
Improved classroom experience through reduced latency
SMART Technologies is a Canadian education technology leader best known for its interactive SMART Boards and digital learning platforms. The company serves schools, businesses, and government agencies across more than 175 countries. Their education software platform, Lumio, enables collaborative, real-time learning across devices and saw rapid growth during the pandemic.
As usage increased, so did demands on their infrastructure, especially for features like real-time whiteboarding, live document editing, image handling, and session management. To continue delivering a seamless classroom experience at scale, SMART required a more powerful and scalable backend.
SMART’s existing architecture was built on Firebase/Firestore, which formed the backbone of their real-time SaaS experience. While effective in earlier stages, the architecture began to strain under rising load and user expectations.
Firebase capped at 20,000 concurrent connections and 1,000 IOPS per instance. Peak traffic routinely hit about 40,000 read/write operations per second, forcing SMART to shard the application across more than ten Google Cloud and Firebase projects. This approach amplified complexity and diluted visibility.
Latency had also become a serious concern. Real-time collaboration experiences—central to Lumio’s value proposition—were being degraded by slow syncs, inefficient image handling, and architectural constraints that made grouped object updates challenging.
The complex data model for the Lumio whiteboarding feature was a primary source of challenges. It frequently caused data inconsistencies, hindered real-time performance, and complicated the implementation of features like image uploads and grouped object manipulation within the existing architecture.
With a 2x year-over-year growth forecast, the system was rapidly becoming unsustainable. SMART needed a new foundation that could scale horizontally, provide better visibility, and deliver real-time performance at a global level.
SMART partnered with OpsGuru, an AWS Premier Consulting Partner, to design and demonstrate an AWS-native backend that could resolve the limitations of their existing Firebase infrastructure. The engagement was centered around re-architecting the core backend for real-time collaboration, ensuring it could meet SMART’s performance targets while simplifying day-to-day operations.
The new architecture was built on AWS using a VueJS frontend with a Java Spring Boot API running on Amazon ECS Fargate. MongoDB Atlas replaced Firestore as the primary NoSQL document store, configured with sharding to ensure future scalability. Redis, delivered via Amazon ElastiCache, handled distributed locks and real-time pub/sub messaging for web socket server-to-client communication.
WebSocket traffic terminates at an Application Load Balancer that sits behind Amazon CloudFront, placing edge nodes close to classrooms worldwide and providing better edge delivery and lower latency. Observability is achieved through integration with SMART’s existing Splunk platform. Performance and resilience were proven using Grafana k6 load scripts executed from multiple AWS Regions.
The new application retained critical capabilities, including collaborative session creation, document editing, and real-time updates, with integrated user authentication via SMART’s existing identity provider. Excalidraw was used as the front-end canvas and whiteboarding solution.
New data models were used to simplify object grouping and image uploading, which enabled integration with SMART’s latest content service. Conflict testing was performed using timed and randomized race-condition generation to ensure the new system could handle collaborative stress scenarios.
To validate the architecture under load, OpsGuru implemented distributed load testing using Grafana k6. The tests confirmed that core services, such as Redis and Fargate, scaled effectively. Initially, MongoDB emerged as a bottleneck, but after resizing the cluster and optimizing shard distribution, the platform achieved a sustained throughput of over 50,000 writes per second, exceeding SMART’s projected needs.

“Working with OpsGuru was a turning point for us. They not only understood the complexity of our platform but delivered a production-grade AWS foundation that exceeded our expectations. We were hitting serious limits with Firebase, but OpsGuru helped us build a scalable, modern backend that comfortably meets our current needs—handling 50,000 writes per second—and is designed to scale well beyond that. The performance gains, architectural clarity, and readiness for global growth we achieved in this engagement have set the foundation for the next chapter of Lumio—our real-time, collaborative learning platform.”
— Mugdha Jain, Director, Software Development at SMART
The AWS prototype proved that SMART could meet its performance and scale goals while simplifying infrastructure management. Redis and Fargate handled load spikes with ease, ensuring seamless real-time communication even under high concurrency. MongoDB, once properly scaled and tuned for SMART’s workload, supported sustained throughput exceeding 50,000 writes per second - more than 2.5 times the original Firebase system’s limit.
These performance improvements will restore the fast, fluid collaboration experience Lumio users expect. In fact, teachers and students will now benefit from a noticeably smoother and more intuitive experience, with latency improvements that make the experience feel close to twice as fluid and intuitive compared to the previous setup.
Beyond raw performance, the shift to AWS brought meaningful operational efficiencies. The Firebase-based system had required SMART’s team to manage and monitor over ten separate Google Cloud and Firebase projects. This not only consumed engineering resources but also introduced complexity and risk.
With the new architecture, SMART consolidated everything into a single, unified deployment, reducing overhead, increasing visibility, and freeing up valuable engineering hours previously spent managing fragmented infrastructure.
More importantly, the new architecture gives SMART room to grow. With horizontal scalability built into the design and a robust, cloud-native foundation, SMART is now well-positioned to onboard new schools, districts, and users without fear of infrastructure failure or saturation. No longer constrained by Firebase’s per-project caps, the team can easily scale usage without reconfiguring backend resources or spinning up new deployments.
This architectural simplicity not only accelerates customer onboarding but also opens the door to deeper integrations with third-party platforms and analytics tools. By centralizing data into a single source of truth rather than sharded data across multiple Firebase projects, SMART can now explore richer insights into classroom usage and product performance and unlock future potential in areas like personalized learning, content optimization, and platform intelligence.
The result is a production-ready architecture that will not only solve today’s performance bottlenecks but positions SMART for continued innovation, scale, and impact in the global EdTech landscape.