What Actually Breaks at 5 Million Requests a Day
Notes from building and hardening an API gateway handling 5M+ requests a day at 99.95% availability: the failure modes that do not show up until real traffic, and the patterns that actually held.
Blog
Notes from production systems
Original writing on architecture, event-driven systems, reliability and engineering careers — drawn from production work on banking and high-traffic platforms, with the numbers and the trade-offs left in.
3 articles
Streaming Pipelines for Banking Billing and Payments: Kafka, NiFi and Ignite
Banking data arrives as files on somebody else's schedule and has to be processed as a stream with money-grade correctness. Why the ingest tool and the event backbone are different tools, and what a distributed cache is genuinely for.
Real-Time Analytics with Kafka Streams, Elasticsearch and Kibana
Building analytics that answer questions about the last five minutes, not last night: windowing and the late-event problem, why the state store needs a changelog topic, and the split between what Kafka Streams should compute and what Elasticsearch should.