Content-Addressed Storage with BLAKE3 for SQL Database Value Deduplication
We present a content-addressed storage (CAS) system integrated into a SQL database engine, using BLAKE3 cryptographic hashing for automatic deduplication of large text and binary column values. Values exceeding a configurable threshold (default 1,024 bytes) are stored in a dedicated CAS keyspace under their BLAKE3 hash, with the original row storing a lightweight CasRef pointer. Duplicate values across rows, tables, and copy-on-write branches share a single physical copy. The system is exposed via a SQL column attribute (STORAGE CONTENT_ADDRESSED) and operates transparently on INSERT/UPDATE/SELECT. While content-addressed storage is used in version control systems (Git: SHA-1/SHA-256) and distributed storage (IPFS: SHA-256, Dolt: prolly trees), no existing SQL database uses BLAKE3 for content-addressed value deduplication or exposes CAS as a declarative SQL column attribute.
Authors
- Daniel Moya Vaca
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-03-27
- DOI
- https://doi.org/10.5281/zenodo.19244285
- Primary Topic
- Cloud Data Security Solutions
- Type
- article
- Field-Weighted Citation Impact
- 0.00