Search Manual / Version 2512.0
Table Of ContentsStoring vector embeddings for semantic search requires significantly more space in the search index than traditional metadata. The following table provides example sizes for float32 vectors (1024 dimensions) per document:
| Documents | Estimated Size |
|---|---|
| 1 | ~4KB |
| 1,000 | ~4–7MB |
| 1,000,000 | ~4–7GB |
Table 4.1. Estimated Storage Requirements for Vectors
With multimodal embedding, both media and text vectors are stored for images. This means two vector fields per image, effectively doubling the storage required compared to plain text documents.
Example: For 2,000 images (each with 2 vectors) and 1,000 articles (each with 1 vector), the total is 5,000 vectors. Using a planning estimate of 8MB per 1,000 vectors, this results in approximately 40MB of storage required for the vectors.
These are rough estimates; actual storage requirements may vary depending on your Solr and Embedding Service configuration.
