content-addressed-storage · shipping · also Content-addressable storage, CAS

Content-addressed storage

Content-addressed storage retrieves data using identifiers derived from its content under specified encoding and hashing rules.

Related Observatory article: Read on hesela.com

A content address names a representation rather than a storage location. Venti used block fingerprints for lookup, duplicate coalescing and read verification. Its lookup key paired the fingerprint with an 8-bit client-supplied type; the server did not interpret that type's data format. Its 2002 SHA-1 assumptions are historical, not a current security recommendation. [4]

Git hashes an object header containing type and byte length, a NUL byte, and the payload. A blob omits its filename; a tree associates names and modes with object IDs. Equal file bytes therefore need not mean equal directory objects. [1]

Chunking chooses the bytes to identify; content addressing names the resulting representation. IPFS root identifiers also depend on chunking, graph layout, codec and hash choices. Identical file bytes alone do not fix one identifier across different import settings. [2]

The acronym CAS also means compare-and-swap, an unrelated conditional-update operation.

Relations

related
Content Identifier (CID), Content-defined chunking, Data deduplication, Cryptographic hash function, Merkle tree, Compare-and-swap

Limitations

  • A retained identifier does not retain the addressed data. IPFS pinning protects against a node's garbage collection, not every cause of unavailability. [3]
  • Deduplication benefits depend on repeated content and system policy; a content address alone gives no physical-space saving or restore-speed estimate. [4]

Citations

  1. Git Book: Git Objects, object storage and tree entries (accessed 2026-11-18)
  2. IPFS documentation: content addressing, root blocks and import parameters (accessed 2026-11-18)
  3. IPFS documentation: persistence, garbage collection and pinning (accessed 2026-11-18)
  4. Quinlan and Dorward. Venti: a new approach to archival storage. FAST 2002, sections 3 and 7. (accessed 2026-11-18)
Sources & caveats
  • Primary-source synthesis; historical Venti properties are not universal requirements of content-addressed systems. No native IPFS or Venti experiment was run.

JSON: /data/terms/content-addressed-storage.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Content-addressed storage. Hesela. 2026-11-18. https://hesela.dev/terms/content-addressed-storage/