Hitting the disk write every 3 seconds!

Question

Hello,
Oracle seems to be hitting the disk every 3 seconds - on checking I find that it is the
oracle checkpoint process which writes to the control files every 3 seconds!
Is there a way to make it longer?
There are no parameters in the init.ora for this one.
(log_checkpoint_interval is 10000 and log_checkpoint_timeout is 1800).
I have tried with Oracle 8.1.6 and 8.1.7 on redhat7 - same results.
The disk access is driving me crazy!
strace on ckpt process:
pwrite(407, "\25\2\0\0\3\0\0\0Q\26\0\0\377\377\1\4\257I\0\0\2\0\0\0"..., 4096, 12288) = 4096
pwrite(406, "\25\2\0\0\3\0\0\0Q\26\0\0\377\377\1\4\257I\0\0\2\0\0\0"..., 4096, 12288) = 4096
pwrite(405, "\25\2\0\0\3\0\0\0Q\26\0\0\377\377\1\4\257I\0\0\2\0\0\0"..., 4096, 12288) = 4096
ls -l /proc/<ckpt-pid>/fd also shows the files being written.

Answer 1

I believe that this is part of the incremental checkpoint process of 8i. Every 3 seconds CKPT scans the checkpoint queues, finds the lowest RBA (redo byte address)of the buffers at the heads of the queues and writes this out to the control file. This is written to a "thread private" location in the control file, hence the call to pwrite. This is a light-weight control file update using an offset into the control file and is very efficient. This recorded RBA represents the current position of the incremental checkpoint for the instance.
(Note that LOG_CHECKPOINT_INTERVAL and LOG_CHECKPOINT_TIMEOUT mean completely different things in 8i than previously.)