Skip to main content
This topic explains how to store data in RisingWave. Whenever you create either a materialized view or a table, RisingWave will persist the internal data inside object store (e.g., Amazon S3). RisingWave offers two types of storages: row-based storage (default), and columnar-based storage (Apache Iceberg).

​
Row-based storage

By default, tables and materialized views are stored in row-based storage, using a specialized storage engine called Hummock. The row-based storage will automatically cache data in local memory in order to get optimal storage. The row-based storage is good at serving ad-hoc point and short-range access. But it’s not optimized for long-range access and full-table scans.

​
Columnar-based storage

RisingWave offers columnar-based storage. Instead of using proprietary format, RisingWave stores analytical data based on Apache Iceberg. For more details, see Iceberg table engine.