Re: Potential autovacuum optimization: new tables

On Sat, Oct 13, 2012 at 3:49 PM, Joshua Berkus <josh(at)agliodbs(dot)com> wrote:
> So, problem #1 is coming up with a mathematical formula. My initial target values are in terms of # of rows in the table vs. # of writes before analyze is triggered:
>
> 1 : 3
> 10 : 5
> 100 : 10
> 1000 : 100
> 100000 : 2000
> 1000000 : 5000
> 10000000 : 25000
> 100000000 : 100000

Do we necessarily care about smoothness?

If we don't at all, then this would be fine:

func powerlaw (tuples int) int {
if tuples < 10 {
return 3
}
if tuples < 100 {
return 5
}
if tuples < 1000 {
return 10
}
if tuples < 100000 {
return 100
}
if tuples < 1000000 {
return 2000
}
if tuples < 10000000 {
return 5000
}
if tuples < 100000000 {
return 25000
}
return 100000
}

If we want smoothness within the ranges, this is a piecewise linear
representation of your table:

func powerlaw2 (tuples int) int {
if tuples < 10 {
return 3
}
if tuples < 100 {
return 5 + 5 * (tuples - 90)/90
}
if tuples < 1000 {
return 10 + 90 * (tuples - 900)/900
}
if tuples < 100000 {
return 100 + 1900 * (tuples - 99000)/99000
}
if tuples < 1000000 {
return 2000 + 3000 * (tuples - 900000)/900000
}
if tuples < 10000000 {
return 5000 + 22000 * (tuples - 9000000)/9000000
}
if tuples < 100000000 {
return 25000 + 75000 * (tuples - 90000000)/90000000
}
return 100000
}

That's in Go, but there shouldn't be anything too unfamiliar looking
about it :-).

It would be nice to have a simpler functional representation, but the
above is by no means heinous, and it's not verbose beyond reason.
--
When confronted by a difficult problem, solve it by reducing it to the
question, "How would the Lone Ranger handle this?"