Databases Reference
In-Depth Information
3.
From Tuesday through Saturday, you take a level 1 differential backup that copies all
changed blocks since the level 1 backup the day before.
4.
If you have to recover the database on a Saturday morning, you'll need the previous
Sunday's level 0 backups plus all the differential incremental level 1 backups from Monday
through Friday.
A cumulative level 1 backup always takes longer than a differential backup, since it backs up all changed blocks
since the last level 0 incremental backup. Thus, cumulative backups need more time as well as space, since they
“repeat” or “duplicate” the copying of changed blocks. Differential backups, on the other hand, don't duplicate the
work performed by previous backups at the same level—a differential incremental level 1 backup done on a given day
is always distinct from the same-level backup done the day before.
You can perform incremental backups of any of the following:
•
Data file
•
Data file copy
•
Tablespace
•
Database
You
can't
perform an incremental copy of a control file, archived redo log, or backup set.
While incremental backups do, in general, take significantly less time to complete than a full backup of the same
files, you can't be absolutely sure that this is always true. This is because of how RMAN checks data blocks for changes.
Even during an incremental backup (at a greater than level 0 incremental backup), RMAN still reads all data blocks
in a data file into the memory to check the block's SCN number. Any block with an SCN more recent than the SCN of
the level 0 incremental backup is moved from the input buffer to the output buffer, and from there it's written to the
backup piece.
■
if you want fast incremental backup performance, use the block change tracking feature, where RMAN doesn't scan
all the data blocks to see whether they've changed, to determine whether they are candidates for the incremental backup.
Note
You can't use an incremental backup directly during a database restore operation since it's only a complement
to a full backup and can't be “restored.” It's only to provide a faster recovery time (faster mean time to recovery,
or MTTR). The following example serves to demonstrate this point:
RMAN> run
{
restore datafile 7;
recover datafile 7;
}
Once RMAN restores data file 7 from the latest level 0 incremental backup, it has two choices. It can use
incremental level backups since the most recent level 0 backup and add any necessary archive logs to recover the
database to the present point in time. Alternatively, RMAN can choose to use archived logs only from the level 0
backup time to recover. RMAN always prefers using incremental backups to archive logs.
