Lessons learned from scaling to 1 million Lambda functions | Amazon Web Services

What this post added

This post details the lessons learned from scaling a serverless SaaS platform to over one million Lambda functions across thousands of AWS accounts. It covers the evolution from initial microservice composition with CloudFormation StackSets to addressing scaling challenges like self-DDoS via request scattering and optimizing multi-account observability costs. A significant contribution is the re-evaluation of SQS for scale-to-zero, leading to its removal from the EventBridge-to-Lambda path and the implementation of metric-driven safety and a centralized DLQ. The post also highlights the journey from building a custom deployment engine to collaborating with the AWS CloudFormation team to industrialize the deployment process at scale.