System Design Interview
High-QPS search autocomplete that returns ranked suggestions under 100 ms with fresh, abuse-resistant, and scalable indexing.
Single service handles suggest logic with in-DB prefix tables.
What was missing: Edge protection, caching, indexing pipeline, and observability.
Why that's risky: Hot rows and no isolation; slow under spikes.
What gets added: Nothing yet (MVP).
Trade-offs: Fast to ship; not production-ready.
Protect ingress and separate compute from persistence.
What was missing: Ingress protection and clear separation of concerns.
Why that's risky: Origin overload and lack of quotas.
What gets added: Edge gateway and rate limiting before API.
Trade-offs: Quota config and tuning overhead.
Scale read path horizontally and add a hot result cache.
What was missing: Horizontal scale and hot-path caching.
Why that's risky: High p99 and DB pressure under spikes.
What gets added: Replica pool behind LB and prefix cache.
Trade-offs: Cache invalidation complexity.
Introduce a search-optimized index and pipeline fed by logs and content updates.
What was missing: Search-optimized storage and async updates.
Why that's risky: DB hot rows and slow writes under load.
What gets added: Dedicated index, queue, indexer, and analytics ETL.
Trade-offs: Eventual consistency of index vs source of truth.
Add ranking service and offline features to improve relevance.
What was missing: Relevance improvements and user-aware ranking.
Why that's risky: Generic results and poor engagement.
What gets added: Ranking service and features store with signal aggregation.
Trade-offs: Feature staleness and complexity in experimentation.
Improve operability and resilience with monitoring, index replicas, and proactive cache warming.
What was missing: Operational visibility and resilience for cache/index.
Why that's risky: Blind spots during incidents; slow warm cache on deploy.
What gets added: Monitoring, index replica, and cache warming worker.
Trade-offs: Extra capacity cost for replicas and warmers.
Use a trie or n‑gram index specialized for prefix lookups. If you also need full‑text features, maintain both: an inverted index for search and a compact trie/n‑gram for suggest. Keep indexes in sync via an async pipeline.
Put a hot result cache in front of the index, shard the suggest replicas behind a load balancer, and keep results small (limit K, compress). Use timeouts/circuit breakers and pre-warm caches on deploy. Track tail latency and cache hit ratio closely.
Combine popularity and recency signals from logs with optional user features (locale, history). Run a Ranking Service that merges index candidates with feature lookups; compute features offline and store in a features DB for low-latency reads.
Stream deltas from content changes and logs into a queue; batch and debounce updates in workers. Maintain a DLQ for bad events and rate-limit backfills. Expose tooling to pause/resume pipelines and rebuild specific shards/sectors.
Maintain a policy filter that runs post-ranking to remove banned terms and low-quality suggestions. Support tenant-specific blocklists and thresholds. Log rejections for review and retraining.
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:
