XIII. Case Studies
Theory becomes conviction only when you see it survive contact with production. This part reads the primary sources — original papers, official postmortems, first-party engineering blogs — of systems that operate at a scale where every trade-off in this handbook has a price tag attached. Each case study connects a real architecture or a real outage back to the mechanisms that produced it, closing the loop between the principles and the systems that live or die by them.
Topics Covered
Section titled “Topics Covered”- 13.1. Messaging Platform: Apache Kafka: The distributed commit log in depth: disk persistence and zero-copy, KRaft mode, consumer rebalancing, and exactly-once semantics.
- 13.2. Distributed Key-Value Store: DynamoDB / Cassandra: The origin of the AP database: the Dynamo paper’s lineage in DynamoDB and Cassandra, their data models, and the hot-partition problem.
- 13.3. Microservices at Scale: Netflix, Uber, Airbnb: How Netflix, Uber, and Airbnb operate microservices at scale, and the postmortem and technical-debt culture that sustains it.
- 13.4. Real-Time Analytics Platforms: Sub-second queries over arriving data: stream joins, Pinot and Druid, the path from Lambda to Data Mesh, and ML serving.
- 13.5. Distributed SQL: Spanner & CockroachDB: Scale and strong consistency together: Spanner and CockroachDB, TrueTime and HLC, and two-phase commit over consensus.
- 13.6. Anatomy of Famous Outages: Distributed-systems lessons from real incidents: the GitHub, Roblox, Cloudflare, and Meta outages, read from their official postmortems.