Sharding Strategies for Vector Search (EXPLAINED)
Horizontal scaling patterns: hash sharding, semantic partitions, routing embeddings, and merge/rerank at query time.
TL;DR — Quick Answer
Shard by tenant, geography, or embedding model version using consistent hashing or dedicated collections, query relevant shards in parallel, then merge/rerank global top-k. Avoid naive random sharding that splits semantic neighborhoods. Plan rebalance for hot shards and cap fan-out to control p99 latency.
The Interview Question
How do you shard vector indexes at scale while preserving recall and manageable query fan-out?
Deep Explanation
Sign in to unlock full answer
Get deep explanations, PDF export & all Vector Databases questions
- 22 more sections of deep explanation
- Real-world examples
- Common mistakes
- Interviewer expectations
- Follow-up questions
ShardingScaleFan-OutArchitectureUberAirbnbPinecone