Reactive Architecture on AWS: from 0 to 1.6M events/minute
summary
Bitpanda Pro scales a crypto exchange from a four‑person team to peaks of 1.6 million events/minute, starting with a Vert.x monolith on ECS Fargate backed by Aurora RDS, then evolving to Kafka/MSK, DynamoDB, SQS and EventBridge; a later hot‑path redesign uses EC2 plus Chronicle for ultra‑low latency. The system handles 10x scale spikes with zero recorded errors at the high watermark, but end‑to‑end P99 100ms latency remains above the sub‑millisecond goal and large historical data migrations prove costly.
AI-written summary · about
chapters (17)
- 0:00Intro
- 2:32Typical use cases
- 4:22N-tier architecture
- 5:42Reactive applications
- 6:34Design principles
- 12:23Challenges
- 15:06AWS-our first take
- 19:52First take - unresolved problems
- 21:06Second take-elasticity & message passing
- 22:23Second take-outsourcing SQL reads
- 24:56Reactive thus far-market volatility
- 26:18Lessons learned
- 28:37Expanding our AWS universe
- 30:19Tiered architecture
- 31:15Navigating (reactive) AWS
- 34:29Drivers of resilient architecture
- 36:09Additional resources
related talks
59mBuilding resilient event-driven architectures
44mIntroduction to application scaling
54mReacting to an Event-Driven World
25mBest practices for advertising and marketing technology on AWS
44mAdvanced Event-driven Architectures at
01h 11mAsychronous design with Spring and RTI: 1M events per second
18m"Kafka's Metamorphosis"
39m400 Billion Messages a Day with Kafka