Design a scalable search service with sharding

Technical Design: Search + Sharding Trade-offs

Design a search service that lets users query documents/products with keywords.

Functional requirements

• GET /search?q=... returns top results ranked by relevance.
• Support basic filtering (e.g., category, language, tenant/app id).
• P95 latency target (e.g., < 200 ms) and ability to scale to large corpora.

Non-functional requirements

• Data size grows to billions of docs.
• Frequent reads, continuous ingestion/updates.
• High availability.

Prompt

1. Describe how you would implement the core search algorithm at a high level (indexing + querying).
2. Propose a sharding strategy for the index.
3. Discuss trade-offs: latency, cost, rebalancing, hot keys, consistency, and operational complexity.

No need to write production code, but be concrete about components and data flows.