Hybrid tables

Snowflake hybrid tables offer an innovative solution to the distinct but conflicting demands of transactional and analytical workloads. While transactional tasks require small, random, and high-concurrency reads and writes, analytics on the same dataset call for large-scale in-memory processing. The dual-storage architecture of hybrid tables addresses these needs by employing row-based storage for fast transactional operations and parallel columnar storage for analytical queries—seamlessly switching according to the workload type and providing optimized performance for each case.

Row-based storage allows hybrid tables to offer features such as row-locking and indexing and employ constraints such as primary and foreign keys while enforcing referential integrity (RI). This unlocks workloads that would otherwise be impractical in a column store architecture such as transactional applications that must maintain state and perform atomic transactions across tables...