no need to undo any writes. Using immediate updates, an update may be made to the database itself any time
after a log record is written. The log can be used to undo and redo transactions in the event of failure.
n Checkpoints are used to improve database recovery. At a checkpoint, all modified buffer blocks, all log
records, and a checkpoint record identifying all active transactions are written to disk. If a failure occurs, the
checkpoint record identifies which transactions need to be redone.
n Advanced transaction models include nested transactions, sagas, multilevel transactions, dynamically
restructuring transactions, and workflow models.