Backed by Relevance Lab· previously known as Kaman.aiInsightsWhitepapersTrust centreCareersContactSign in
Try it now

Home / Whitepapers / Whitepaper

Whitepaper

KDL: an embeddable, governed lakehouse engine speaking the Postgres wire protocol

A description and measurement of KDL, Camaan's lakehouse engine. It stores data as Parquet files, takes small durable writes through a group-commit log, applies row policies and column masks inside the query plan, and serves clients over the Postgres protocol.

  • Governance inside the engine: row policies and column masks are query-plan rewrites bound to the signed-in person.
  • One write path for streams and single rows, with group commit sharing a disk sync across writers.
  • A measured latency floor: a single-writer commit at a median 2.5 ms on NVMe, and a model that predicts write scaling to within a few percent.
  • External tables, change-data capture from a message stream, spatial SQL and in-engine forecasting.
KDL: an embeddable, governed lakehouse engine speaking the Postgres wire protocol | Camaan