6 years of serverless – Lessons learned from real-world applications
summary
Grohe Sense/Guard runs on AWS serverless stack for six years, handling 288 million MQTT messages and about 1.6 billion Lambda invocations per month. Architecture uses IoT Core, Lambda, DynamoDB, S3, SNS, SQS and Step Functions; everything-as-code, heavy monitoring, and domain-owned data. Key lessons: avoid long-running Lambdas and synchronous Lambda chains, prefer push/queue patterns, design idempotent operations, minimize cold-starts (small packages, provisioned concurrency costs money), and use Step Functions for retries, delays and complex workflows.
AI-written summary · about
chapters (15)
- 0:00Intro
- 0:56Grohe SenseGuard Real-time water waste and damage prevention
- 4:56Traits of serverless
- 16:25loT temperature sensor
- 23:29Use Lambda for long running calls
- 25:12Push, don't pull
- 27:33Triggers
- 30:02Understanding AWS Lambda scaling
- 39:55Step Functions for Workflows
- 41:53Heavy start ups
- 44:11Consider Lambdas domain boundaries
- 45:54Expect errors
- 47:04Aim for idempotency
- 51:13Integration tests
- 54:37Takeaways
related talks
29m2 Years, 0 Servers: Lessons learned from running a startup 100% on Serverless
17mHow do we use Serverless Architectures on AWS Lambda
40m4 Key takeaways from running Serverless on production for 4 years
42mNo Server Is Better Than No Server: Building Blocks for Serverless Architectures
45mPatterns & practices for developing enterprise-scale serverless applications
39mBuilding Highly Scalable Microservices with Serverless
23mFrom 1 to 101 Lambda Functions in Production: Evolving a Serverless Architecture–Slobodan Stojanovic
01h 00mServerless - reality or BS - notes from the trenches