Learning path

Patterns

Realtime, async, scale, contention

  1. Common system design patterns The recurring system design interview patterns — realtime updates, long-running tasks, contention, scaling reads and writes, large blobs, multi-step workflows, and proximity — with the trade-offs and failure modes behind each.
  2. Realtime updates The real-time updates pattern end to end: two hops (client protocols + server fan-out), networking basics, polling vs long poll vs SSE vs WebSockets vs WebRTC, L4/L7 load balancers, consistent hashing vs pub/sub, deep dives, and when to stay simple.
  3. Long-running tasks Split slow work from HTTP: ack with a job ID, durable queues, worker pools, DLQs, idempotency, backpressure, and mixed workload queues — with interview scenarios for YouTube, Instagram, Uber, and Stripe.
  4. Multi-step processes & sagas Reliable multi-step workflows end to end: single-server pitfalls, sagas and compensation, event-driven choreography vs orchestration, Temporal durable execution, Step Functions, signals, idempotency, versioning, and interview deep dives.
  5. Scaling reads Handle massive read load without crushing the primary: indexes and denormalization, read replicas and sharding, Redis and CDN caching, then hot keys, cache stampede, and versioned invalidation — with interview scenarios.
  6. Scaling writes Handle high-volume writes when one database melts: vertical scale and write-optimized stores, sharding and vertical partitioning, queues and load shedding, then batching and hierarchical aggregation — with hot keys, resharding, and interview scenarios.
  7. Dealing with contention Race conditions, lost updates, and the coordination ladder: conditional writes, pessimistic locking, optimistic concurrency, isolation levels / write skew, distributed locks, and when to serialize — with interview scenarios and deep dives.
  8. Handling large blobs Move videos and documents off your API path: presigned uploads, CDN signed downloads, multipart resume, event-driven state sync, and when not to use direct-to-storage — with interview scenarios for YouTube, Instagram, and Dropbox.

Lattice