WriteAheadLog

Write-Ahead Logging (WAL) is a logging technique used in databases and other data-management systems to ensure atomicity and durability of transactions. The central idea is that all modifications to the database are recorded in a log before they are applied to the data files. By persisting the log first, the system can recover to a consistent state after a crash by replaying or undoing operations described in the log.

Mechanism: For each update, the system creates a log record describing the change and writes this record to stable storage before the corresponding data pages are modified in memory. The log is typically append-only and organized into segments. A commit record may be written and flushed to signal transaction end. Data pages are then updated and may be flushed later, often at checkpoints. This approach allows in-memory changes to be reordered without sacrificing durability.

Recovery: On startup after a crash, the system runs a recovery procedure that analyzes the log to identify committed and uncommitted transactions. It redoes the changes of committed transactions and undoes any effects of uncommitted ones to bring the database to a consistent state as of the last commit. Checkpoints bound the scope of recovery work.

Adoption and variants: WAL is central to many modern relational databases, notably PostgreSQL, which uses a WAL for crash recovery and replication. In MySQL, InnoDB uses a redo log with WAL-like properties. Some databases offer a WAL mode to improve concurrency, with trade-offs in log management and disk I/O.

See also: Journaling, Crash recovery, Durability.