Physical logging, checkpoints, and fast recovery

These topics cover the three procedures that the database server uses to achieve data consistency:
  • Physical logging
  • Checkpoints
  • Fast recovery

The physical log is a set of disk pages where the database server stores an unmodified copy of the page called a before-image. Physical logging is the process of storing a before-image of a page that the database server is going to change. A checkpoint is a point when the database server synchronizes the pages on disk with the pages in the shared-memory buffers. Fast recovery is an automatic procedure that restores the database server to a consistent state after it goes offline under uncontrolled conditions.

These procedures ensure that multiple, logically related writes are recorded as a unit, and that data in shared memory is periodically made consistent with data on disk.

For the tasks to manage and monitor the physical log and checkpoints, see Manage the physical log.