14 articles in high level design › case studies.
How Discord rewrote a critical service from Go to Rust to eliminate GC pauses, achieving consistent sub-millisecond p99 latency and lower memory usage.
How Discord migrated from Apache Cassandra to ScyllaDB to eliminate latency spikes caused by JVM garbage collection, achieving 2-5x throughput improvement at lower cost.
How Slack redesigned its message storage from a flat text log with metadata columns to a relational channel model, enabling more efficient queries and reducing read amplification.
How a single misconfigured BGP update command took Facebook, Instagram, and WhatsApp offline for 6+ hours and locked employees out of the buildings needed to fix it.
How Netflix pioneered intentional failure injection in production to build genuinely resilient distributed systems, and the principles behind Chaos Monkey, Simian Army, and ChAP.
How Stripe's idempotency key design prevents duplicate charges, handles network failures gracefully, and provides a blueprint for safe financial API design.
How Google's Spanner database uses GPS-synchronized atomic clocks and bounded uncertainty to provide globally consistent transactions across data centers on different continents.
How GitHub moved from a single master MySQL cluster to Vitess, enabling horizontal sharding at scale while maintaining MySQL compatibility for application code.
How Airbnb migrated from a Ruby on Rails monolith to a service-oriented architecture, what it took to extract services safely, and the lessons from a decade-long process.
How Notion migrated from a single Postgres database to horizontally sharded clusters to handle explosive growth, and the infrastructure changes required to make it work transparently.
How Twitter evolved from a single Rails/MySQL app to a hybrid push/pull timeline architecture with Redis, Manhattan, and FlockDB, eliminating the fail whale by solving fan-out at scale.
How a configuration error at a small ISP created a BGP route leak that caused Cloudflare, Amazon, and hundreds of other services to become unreachable for 6 hours in 2019.
How Figma built real-time collaborative design on top of a Rust-based operational transformation engine, scaling to millions of simultaneous editors with sub-100ms sync latency.
How Amazon's largest AWS region has experienced multiple large-scale cascading failures, and what the architectural patterns behind each reveal about distributed system failure modes.