Re: Auto-tuning work_mem and maintenance_work_mem

On Wed, Oct  9, 2013 at 06:20:13PM +0200, Pavel Stehule wrote:
>     On Wed, Oct  9, 2013 at 05:01:24PM +0200, Pavel Stehule wrote:
>     >     FYI, this auto-tuning is not for us, who understand the parameters
>     and
>     >     how they interact, but for the 90% of our users who would benefit
>     from
>     >     better defaults.  It is true that there might now be cases where you
>     >     would need to _reduce_ work_mem from its default, but I think the new
>     >     computed default will be better for most users.
>     >
>     >
>     >
>     > then we should to use as base a how much dedicated RAM is for PG - not
>     shared
>     > buffers.
> 
>     Yes, that was Josh Berkus's suggestion, and we can switch to that,
>     though it requires a new GUC parameter, and then shared_buffers gets
>     tuned on that.
> 
>     I went with shared_buffers because unlike the others, it is a fixed
>     allocation quantity, while the other are much more variable and harder
>     to set.  I figured we could keep our 25% estimate of shared_buffers and
>     everything else would fall in line.
> 
> 
> I understand, but your proposal change a logic to opposite direction. Maybe
> better is wait to new GUC parameter, and then implement this feature, so be
> logical and simply understandable.

OK, I can easily do that.  What do others think?

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + Everyone has their own god. +