System Design Interview
Quora-like Q&A platform with feeds, ranking, search, and moderation.
Model questions/answers and basic read pages.
What was missing: No caching, search, ranking, moderation, or async processing.
Why that's risky: Hot pages overload DB; writes coupled to heavy post-processing.
What gets added: Baseline API + DB.
Trade-offs: Simple MVP.
Add caches for hot reads and search for discovery.
What was missing: Discovery and hot-read acceleration.
Why that's risky: DB becomes the bottleneck and search can’t scale.
What gets added: Cache + search index.
Trade-offs: Cache invalidation and eventual consistency for index.
Decouple ranking/moderation/indexing using async events.
What was missing: Async pipelines for ranking/moderation.
Why that's risky: Synchronous ranking adds latency & fragility.
What gets added: Event stream + workers + feed store.
Trade-offs: Eventual consistency; more infra.
Split core product flows into specialized services and introduce a shared event backbone.
What was missing: Specialized services for search, graph relationships, and shared async data movement.
Why that's risky: A single API tier becomes tangled with ranking, search, and follow-graph logic, slowing feature development and scaling.
What gets added: Dedicated question/answer/feed services, discovery + graph services, a graph database, and an event bus.
Trade-offs: More moving pieces, more eventual consistency, and more operational coordination.
Layer on rate limiting, observability, multi-region failover, and backups around the production service graph.
What was missing: Operational guardrails and resilience around the specialized production service graph.
Why that's risky: Without rate limits, telemetry, replication, and backups, the system is fragile during growth, abuse, or regional failures.
What gets added: Global routing, regional edge, observability, rate limiting, primary/replica content DBs, and DR backups around the domain services.
Trade-offs: The design is far more realistic but noticeably more complex to operate and reason about.
Use a mix of quality signals (votes, author trust, freshness) with offline feature pipelines and online scoring; cache the ranked list.
Emit events for changes; indexer applies tombstones; periodic reindex jobs reconcile divergence.
You're in the middle of an interview session. Leaving now will end your current attempt.
Explore concept overviews, real-system examples, key tradeoffs, and interview talking points for each roadmap section.
You've conquered this phase. These are the skills you now own:
