Re: Cost-Based Vacuum Delay tuning

Guillaume Cottenceau wrote:

> I have noticed that others (Alvaro, Joshua) suggest to set
> vacuum_cost_delay as low as 10 or 20 ms,

My suggestion is to set it as *high* as 10 or 20 ms. Compared to the
original default of 0ms. This is just because I'm lazy enough not to
have done any measuring of the exact consequences of such a setting, and
out of fear that a very high value could provoke some sort of disaster.

I must admit that changing the vacuum_delay_limit isn't something that
I'm used to recommending. Maybe it does make sense considering
readahead effects and the new "ring buffer" stuff.

--
Alvaro Herrera http://www.amazon.com/gp/registry/CTMLCN8V17R4
"La experiencia nos dice que el hombre peló millones de veces las patatas,
pero era forzoso admitir la posibilidad de que en un caso entre millones,
las patatas pelarían al hombre" (Ijon Tichy)