fawn-ds · research · also FAWN Data Store

FAWN-DS

FAWN-DS is the log-structured per-node key-value datastore used by the FAWN-KV research system.

A DRAM hash index points into a flash log containing full keys and values. Matching a short key fragment only selects a candidate: lookup compares the complete stored key and continues searching on a mismatch. The six-byte index slot contains a 15-bit fragment, one valid bit and a four-byte log offset. [1]

The complete key here is the 160-bit lookup key. Comparing it cannot distinguish original objects already mapped to the same key upstream; this interface limit is not a measured collision result. [1]

Relations

related
SILT, Key-value separation, Value log

Limitations

  • The datastore interface does not provide relational operations or transactions. [1]

Citations

  1. Andersen, Franklin, Kaminsky, Phanishayee, Tan and Vasudevan. FAWN: A Fast Array of Wimpy Nodes. SOSP 2009, section 3.3 and Figure 3. (accessed 2026-11-07)

Further reading

  1. Hesela: compact flash indexes and explicit slot-occupancy accounting. (accessed 2026-11-07)
Sources & caveats
  • Historical source review, not a current product benchmark. FAWN-DS names the datastore, not the complete replicated FAWN-KV system.

JSON: /data/terms/fawn-ds.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. FAWN-DS. Hesela. 2026-11-07. https://hesela.dev/terms/fawn-ds/