Architecting for Hyperscale with AWS: Navigating Real-world Serverless Challenges
summary
Architecture used for a COVID‑scale launch handles millions: front end on S3+CloudFront, API Gateway (default 10,000 RPS API Gateway), Lambda with default 1,000 Lambda concurrency and new burst model (1,000 containers per 10s per function), RDS Proxy for connection pooling, DynamoDB on‑demand with DAX and partition pre‑warming, ElastiCache, and queueing to decouple. Demonstrated tools: Lambda Power Tuning halving CPU‑bound latency (128→256 MB dropped 2s→1s), provisioned/reserved concurrency and an open‑source virtual waiting room tested at 100k users / 10k RPS.
AI-written summary · about
chapters (18)
- 0:00Introduction and agenda
- 2:05Real-world hyperscale use case
- 3:09API Gateway scaling limits
- 4:30AWS Lambda concurrency explained
- 11:10Provisioned concurrency for Lambda
- 12:55Optimizing Lambda function performance
- 17:54Lambda Power Tuning demo
- 21:56Asynchronous processing with queues
- 25:25Scaling Amazon Aurora databases
- 28:16DynamoDB scaling strategies
- 31:04Multi-layer caching techniques
- 35:39Estimating service quotas
- 39:53Load testing best practices
- 41:18Observability with AWS X-Ray
- 43:25Error handling and retries
- 44:50Virtual waiting room solution
- 45:54Summary of best practices
- 49:04Additional resources and Q&A
related talks
29mBuild highly scalable serverless applications on AWS
49mServerless Architectural Patterns and Best Practices
47mAdvanced Serverless Architectural Patterns on AWS
31mAccelerate business innovation with serverless applications
54mServerless Architectures: no servers allowed!
50mServerless Architecture Patterns and Best Practices
21mBeyond the Basics: Production Serverless Patterns for Extreme Scale
23mSimplifying design complexities using Serverless Architecture on AWS