rowstore

Rowstore refers to a database storage layout in which data is kept row by row, with all attributes of a record stored together in each row. In row-oriented databases, a single disk page or memory block contains most or all fields of a row, and new rows are appended or updated as whole units. This contrasts with columnar stores, which place values from the same column of many rows contiguously to improve column-wise compression and analytical processing.

Rowstore is well suited to online transaction processing (OLTP) workloads, where applications retrieve or modify complete records by primary key or index and where the cost of scanning many columns per row is offset by fast row access. It supports simple row-level locking, efficient concurrent inserts and updates, and natural handling of relational schemas with wide rows.

However, row-oriented layouts can be less efficient for analytical queries that scan large numbers of rows but only a few columns, because reading many unused fields increases I/O and memory use. Data compression tends to be less effective than with column stores, and vectorized execution is harder, reducing cache locality on wide rows.

Modern database systems may offer rowstore as the default storage and also provide columnstore or hybrid HTAP capabilities, enabling workloads that mix transactional and analytical processing. Hybrid approaches allow recent rows to be stored in a row-oriented format for fast writes, while historical or aggregated data are stored columnar for efficient analytics.