Oracle
would dynamically vary the checkpoint frequency in an attempt to limit the number
of blocks that will need to be read for recovery to the value of this parameter.
Oracle9i further sped this recovery process. Beginning at the last checkpoint, the
redo log was scanned for data blocks that contain unsaved changes and need to be
recovered. In the subsequent scan, changes are applied only where needed. Because
the subsequent scan is a sequential read and reading unnecessary blocks (random I/O)
is eliminated, the recovery time is reduced.
Oracle9i introduced an important fast-start time-based recovery feature. DBAs
specify a target for recovery time in seconds (in the FAST_START_MTTR_TARGET
initialization parameter, where MTTR stands for Mean Time to Recover) in order to
meet Service Level Agreements and other requirements. The database automatically
System Failure | 261
determines values for FAST_START_IO_TARGET and LOG_CHECKPOINT_
INTERVAL. Estimated MTTR values are calculated and placed in V$INSTANCE_
RECOVERY, thereby providing a means for real-world calibration and more accurate
estimates over time.
Pages:
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573