Re: Query palns and tug-of-war with enable_sort

--- On Thu, 19/2/09, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> Also, it'd be worth revisiting the question of whether
> you really still
> need enable_sort off ... personally, I'd think that
> reducing
> random_page_cost is a much saner way of nudging the planner
> in the
> direction of preferring indexscans.
>

We have relatively quick storage and most of our data fits in ram, so I've dropped random_page_cost a little more and at some point I'll flick enable_sort back on and see how it goes.

> BTW, it might be a bit late for this, but you'd be a
> lot better off
> performance-wise with bigint join keys instead of
> numeric(8,0).
> Numeric is slow, and at that field width it's not
> buying you anything at
> all.
>

This may be a little out of my control, there's a lot of things wrong with how our tables are set up and I generally have to swim through lots of 20+ year old code to discover how changes will affect it. That said there's a lot of these numeric(8,0) fields and I doubt switching them for bigint would cause any problems.

Thanks Tom.