Re: Logical to physical page mapping

On 10/28/2012 10:50 PM, Peter Geoghegan wrote:
> On 28 October 2012 22:35, Jan Wieck <JanWieck(at)yahoo(dot)com> wrote:
>> The amount of WAL generated with full_page_writes=on is quite substantial.
>> For pgbench for example the ratio 20:1. Meaning with full_page_writes you
>> write 20x the amount you do without.
>
> Sure, but as always, pgbench pessimises everything by having writes
> follow a uniform distribution, which is completely unrepresentative of
> the natural world. This will literally maximise the number of pristine
> full page images that need to be included. The actual frequency of
> checkpoints is a crucial factor here too, and you didn't mention
> anything about that.

Higher buffer cache hit rates certainly reduce that ratio.

Well, I guess it was just one of those random thoughts that can't work
in the end or aren't worth the work anyway.

Jan

--
Anyone who trades liberty for security deserves neither
liberty nor security. -- Benjamin Franklin