Re: Checkpoint question

On 21/07/2003 07:30 u15074 wrote:
> To improve the performance of PostgreSQL, resp. to avoid too frequently
> occuring
> checkpoints, it is recommended to increase the number of wal files
> (checkpoint_segments).
> So I increased the number of checkpoint segments to be able to write a
> lot of
> data into the database without interruptions due to checkpoints.
> My question now is concerning the cost of a checkpoint if I have a lot of
> checkpoint segments. At checkpoint time will there be checkpointed all
> the
> information contained in all log files, lets assume that I have 40
> checkpoint
> segments for example, or will only some informations of the log files be
> checkpointed?
> If all log files are checkpointed, won't this take a lot of time, and if
> so
> wouldn't it be better to leave the number of checkoint segments to a
> smaller value?
> I'm not quite familiar with the checkpoint concept, so maybe I'm wrong
> with my
> thoughts. Can anyone give advice/explanations?

I'm sure there must be some point where increasing the number of wal files
does not improve performance and your reasoning seem right to me. ISTR
reading something in the archives about someone doing some performance
tests with number of wal buffers and not seeing a big difference even when
large numbers were used.

HTH

--
Paul Thomas
+------------------------------+---------------------------------------------+
| Thomas Micro Systems Limited | Software Solutions for the Smaller
Business |
| Computer Consultants |
http://www.thomas-micro-systems-ltd.co.uk |
+------------------------------+---------------------------------------------+