Re: cost_nonsequential_access()

Manfred Koizar <mkoi-pg@aon.at> writes:
> On Tue, 08 Jun 2004 11:13:43 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> This patch seems to do considerably more violence to the equations than
>> is needed to cover that oversight, though.  The old behavior was
>> intentionally nonlinear in relsize; this is not.

> The comment says "entirely ad-hoc" and I didn't see a particular reason
> why the lower half should be nonlinear in relsize while the upper half
> is linear in 1/relsize.

Incremental changes in the relsize fraction are not going to change the
cost much except near 1, so I was after a curve that went like this
(pardon the crude artwork):


rpc                                            ***********************
                                        *******
                                    ****
                                  **
cost                             *
                                *
                               *
                             **
                         ****
                 ********
1.0 *************
    0                           1                                large
                          relsize fraction

I don't think replacing the lower half of this with a straight line
segment is an improvement.

Possibly the relsize axis ought to be measured on a log scale, or
something like that, but that didn't seem to work nicely when relsize
approaches zero.

            regards, tom lane