Thread: Re: [GENERAL] Multiple Indexing, performance impact

Re: [GENERAL] Multiple Indexing, performance impact

From
Tom Lane
Date:
Peter Eisentraut <peter_e@gmx.net> writes:
> No.  I'm concerned that PostgreSQL should work out of the box for
> everyone.

Agreed.

> And I would prefer that PostgreSQL works the same on every
> platform out of the box.

Well, I'm not sure that we need to take that as far as saying that
default NBuffers can't vary across platforms.  It's not like we're
adding or subtracting functionality.  All I want is to have the default
setup tuned a little better than it is now.

> Obviously we've already lost this on systems
> where the default shmmax is 512kB (SCO OpenServer, Unixware) or 1 MB
> (Solaris), and reducing the parameters is clearly not an option.  But if a
> plurality of systems have the default set at 4 MB or 8 MB then we should
> stop there so we don't upset a large fraction of users.

Making sure that default NBuffers stays under the platform's default
SHMMAX would accomplish that goal at least as well, probably better
than trying to have a one-size-fits-all default; especially if we've
already failed to do the latter.

> Btw., do we have any data on how appropriate wal_buffers = 8 is?

Not that I've seen.  It looks like a rather ad-hoc choice to me...
        regards, tom lane