During a checkpoint all writing transactions
will be blocked.
So especially for OLTP environments the goal is to
keep the checkpoint duration short.
HighEnd OLTP system normally have 3 seconds
max. checkpoint duration.
You might consider lowering your LRU_MIN_DIRTY/LRU_MAX_DIRTY
parameters to achieve this goal.
Values of 1/2 or 5/10 for these parameters are quite normal.
There are also other factors which influence the frequency of
checkpoints like the CKPTINTVL onconfig-parameter and the
size of the physical log (if it is 75% full, a checkpoint is
automatically initiated).
If you are running on IFMX 9.2x, you have so called "fuzzy checkpoints". During a fuzzy checkpoint not all dirty pages
are written from the buffer cache to disk, thus resulting in
shorter checkpoint durations.
This feature is not available in IFMX 7.x.
Also the number of buffers, the I/O throughput, kaio etc.
could influence checkpoint durations.
If you really want the best, you might consider ordering
an IFMX consultant to tune your environment.
HTH.
Best regards
Eric
--
IT-Consulting Herber
WWW:
http://www.herber-consulting.de
Email:
eric@herber-consulting.de
***********************************************
Download the IFMX Database-Monitor for free at:
http://www.herber-consulting.de/BusyBee
***********************************************