WriteAheadLogging

Write-Ahead Logging (WAL) is a technique used in databases and some file systems to ensure durability and atomicity of writes. In WAL, all modifications are first recorded in a log before they are applied to the actual data files. The log is written sequentially and durably, and only then are the data pages updated. This arrangement allows the system to recover to a consistent state after a crash by replaying the log to redo committed transactions and to undo any uncommitted work.

Operations are described by log records that capture the intent and details of changes. The log is flushed to stable storage before the corresponding data pages are updated. On transaction commit, a commit record is written and flushed, guaranteeing durability. If a crash occurs, recovery scans the WAL to reapply necessary changes and to roll back incomplete transactions, restoring atomicity and consistency.

WAL is a core component in many database systems, including PostgreSQL, and is also used in SQLite’s write-ahead mode. Variants exist, such as physical WAL, which records changes to data blocks, and logical WAL, which records higher-level operations for purposes like logical decoding and replication. WAL also enables robust replication by shipping log records to standby instances.

Trade-offs include increased write-ahead latency due to the need to persist log records before data pages are updated, but this is typically offset by improved crash resilience and the ability to optimize data-page I/O through sequential logging. Recovery performance depends on the length and structure of the WAL, and some systems offer both physical and logical logging to support both durability and replication needs.