Tuning I/O - Oracle Database 11gR2 Performance Tuning - page 425

The second query shows the number of data blocks written by DBWR and the number of
checkpoints. This is useful for monitoring the number of blocks written by DBWR during
checkpoints.
In step 7, we view the actual value of parameters, that influence checkpoints, obtaining
the following output:
We have seen the use of the LOG_CHECKPOINTS_TO_ALERT parameter in step 2 and step 3.
You can set LOG_CHECKPOINT_INTERVAL to the maximum number of redo log
blocks—equivalent to the size of O/S file blocks. It will be left in the redo log before
a checkpoint occurs. This value cannot exceed 90 percent of the number of redo blocks
that can be stored in the smallest redo log file, to ensure that there won't be log switch
between checkpoints.
The LOG_CHECKPOINT_TIMEOUT is used to set the maximum number of seconds for
which a dirty block can stay in the memory before it's written to disk. The default value
is 1800 seconds. In step 8, we show other initialization parameters related to Fast-Start
checkpointing. The output of this operation is shown in the following screenshot:
Fast-Start checkpointing is configured to assure that the instance recovery time is acceptable;
this target can be achieved by setting one of the parameters, as shown in the previous image.
We can set the FAST_START_MTTR_TARGET parameter to the expected Mean Time To
Recover , that is, the number of seconds required to recover the instance after a crash.
When you set FAST_START_MTTR_TARGET , you cannot use the parameters
LOG_CHECKPOINT_INTERVAL and LOG_CHECKPOINT_TIMEOUT , shown earlier.