Checkpoint
A checkpoint flushes in-memory dirty pages to disk and records the log position, narrowing the log range that crash recovery must replay.
Purpose
To optimize crash recovery time, balancing checkpoint overhead against recovery performance.
Mechanism
At checkpoint time: 1) pause or throttle transactions; 2) flush dirty pages to disk; 3) write a checkpoint log record (LSN). Recovery replays from the most recent checkpoint LSN. Fuzzy checkpoints allow concurrent flushing.
Use cases
- Crash recovery optimization
- Database tuning
- WAL management