Storage considerations for vector indexes

A vector index consumes storage separately from its base table. The storage a vector index uses is determined by the following factors:

• Vector data – Each indexed item stores its vector as 32-bit floating point (f32) values. Storage for the vector portion of an item scales with the number of dimensions in the index. For example, a 1,536-dimension vector uses roughly four times the vector storage of a 384-dimension vector.

• Projected attributes – The projection you choose controls which base table attributes are copied into the index. KEYS_ONLY stores the least data, INCLUDE stores the key attributes plus the non-key attributes you name, and ALL stores every attribute. A broader projection increases index storage.

• Number of indexed items – Only items that contain a valid vector attribute (and, if the index defines a partition key, that partition key attribute) are replicated to the index. Items that are missing the vector attribute do not consume vector index storage.

To control vector index storage, choose the smallest number of dimensions that meets your relevance needs and project only the attributes your application reads directly from search results. For current storage pricing, see the Amazon DynamoDB pricing on the AWS website.