grokking-system-design-interview

Follow a repeatable 7-step framework: clarify functional and non-functional requirements, estimate scale, define the API, design the data model, sketch high-level architecture, deep dive into one component, and discuss bottlenecks and trade-offs.

Core building blocks include caching with Redis, load balancing with NGINX, sharding via consistent hashing, token bucket rate limiting, and message queues with Kafka. Learning these patterns lets you compose answers to any design question.

Fan-out on write pre-computes feeds for fast reads but slow writes, suiting users with few followers. Fan-out on read computes feeds at query time, better for celebrities with millions of followers. Production systems typically use a hybrid approach.

Consistent hashing maps servers and keys onto a hash ring using virtual nodes, routing each key to the next clockwise node. Adding or removing a shard only remaps roughly 1/N of keys, minimizing data movement during scaling.

Choose eventual consistency for social features like likes and follows where temporary staleness is acceptable and availability matters most. Reserve strong consistency for operations like payments where correctness cannot be compromised during network partitions.