MotherDuck organizes data access through organizations, shares, and Ducklings (compute instances). Databases can be shared within organizations or externally via share URLs, creating read-only zero-copy clones due to DuckDB's single-writer constraint. Read Scaling addresses performance bottlenecks by automatically spinning up additional read-only Ducklings (up to 16 by default) when multiple users query shared data simultaneously, particularly useful for BI dashboards. The system maintains session affinity through session hints and supports eventual consistency, with replicas syncing within minutes. Future DuckLake-based storage will enable multi-account write access with append-only concurrent writes.
Sort: