"physical writes non checkpoint" is low, but increase redo doesn't help.

Question

We have a very busy database, the physical write is so high that we want to reduce the physical write. From the statspack, we could see "physical writes non checkpoint" is low in comparison with "physical write", so I think the physical write resulted from checkpoint is high, then I increased the redo size from 2G to 4G (now the log file switch every 4 minutes). I hope this could reduce the physical write, but the fact is the increased redo is helpless.
Statistic                                      Total     per Second    per Trans
physical writes                            8,298,681        9,210.5        249.8
physical writes non checkpoint             3,229,321        3,584.2         97.2We are using oracle 10.2.0.3 on solaris 10, parameters related to checkpoint are
fast_start_io_target                          0                 
fast_start_mttr_target                        0                 
log_checkpoint_interval                       5000000           
log_checkpoint_timeout                        0     Thanks,
Edited by: user646745 on Aug 18, 2009 1:24 AM

Answer 1

Your log_checkpoint_interval is still at 2.5GB. So, even though you have increased the redo log file sizes, a checkpoint will occur at every 2.5GB of Redo. The rate of checkpointing has moved from "every 2GB of redo" to "every 2.5GB of redo".
You could set log_checkpoint_interval to 0 to disable the intermediate checkpoints.
http://download.oracle.com/docs/cd/B19306_01/server.102/b14237/initparams109.htm#i1129193
Note : However, FAST_START_MTTR_TARGET will still cause Incremental Checkpoints.