> > because the cache will be emptied by high priority multiple new rows,
> > thus writing to the end anyways.
>
> Yes, but this only happens when you don't have enought spare idle CPU
> time. If you are in such situation for long periods, there's nothing you
> can do, you already have problems.
I think such a process would not need a lot of CPU, but a lot of IO. The common
schedulers do not take IO into account (at least not in the here needed sense),
thus you cannot use the process priority mechanism here :-(
An idea could be to only fill the freepage cache from pages that currently reside in the
page buffer. This would also somehow improve cache efficiency, since pages
that are often accessed would get a higher fill level.
A problem with this is, that an empty page (basically an optimal candidate for the list)
would not get into the freelist unless somebody does a seq scan on the table.
Andreas
