A catalog of 120 million items needs vector search where storage cost is the main pressure and some extra build effort is acceptable, with strong ranking. Which configuration fits?
- AA relational index holding no vectors
- BA storage-optimized index with hybrid search and reranking
- CA standard index sized for low latency at any cost
- DA single whole-catalog chunk embedded once
At 120 million cost-sensitive items with acceptable build effort, a storage-optimized index with hybrid search and reranking fits. Standard indexes cost more at scale; the others are not vector search.
Why each option is right or wrong.
- B (correct). Storage-optimized suits huge cost-sensitive corpora, with hybrid search and reranking for quality.
- A. A relational index has no vector search.
- C. A standard index costs more at 120 million items.
- D. One whole-catalog chunk defeats retrieval.
Reference: GenAI Exam Guide Domain 4 (Vector Search configuration); Databricks docs.