Incremental checkpoints are only written to the controlfile, not all the
datafile headers. Nevertheless, FAST_START_IO_TARGET (or
DB_BLOCK_MAX_DIRTY_TARGET under 8.0) should not normally be set unless you have
a very large cache. The reason is that it makes DBWn work harder than necessary.


-----Original Message-----
Sent: Tuesday, 5 September 2000 6:09


Thanks for your help with my other 2 questions. Its much appreciated. Just one
more if you have time!
In 8i Fast Start Checkpointing was introduced so that DBWR processes continually
write buffers to disk to advance the checkpoint position. The smaller the value
of FAST_START_IO_TARGET the more aggressive the checkpointing will be. Any
subsequent crash/instance recovery will be quick.
Will significant overhead result with Fast Start Checkpointing during normal
processing, bearing in mind that the CKPT process must update the headers of all
datafiles to record details of the checkpoint?