A CTO is evaluating vector databases for an enterprise RAG (Retrieval-Augmented Generation) system. Which indexing algorithm provides the best balance of query speed and recall accuracy for high-dimensional embedding search at billion-scale?
-
A
B-tree indexing with range partitioning
-
B
Hierarchical Navigable Small World (HNSW) graphs
-
C
Inverted file index (IVF) with flat quantization only
-
D
Full sequential scan with cosine similarity