Re: Scaling shared buffer eviction

On Thu, Sep 25, 2014 at 9:14 AM, Andres Freund <andres@2ndquadrant.com> wrote:
>> Why stop at 128 mapping locks?   Theoretical downsides to having more
>> mapping locks have been mentioned a few times but has this ever been
>> measured?  I'm starting to wonder if the # mapping locks should be
>> dependent on some other value, perhaps the # of shared bufffers...
>
> Wrong way round. You need to prove the upside of increasing it further,
> not the contrary. The primary downside is cache hit ratio and displacing
> other cache entries...

I can't do that because I don't have the hardware.  I wasn't
suggesting to just set it but to measure the affects of setting it.
But the benefits from going from 16 to 128 are pretty significant at
least on this hardware; I'm curious how much further it can be
pushed...what's wrong with trying it out?

merlin