Re: Performance impact of record sizes

Tom Lane wrote:
> There is no savings from using CHAR(n) --- most of the time it's a
> substantial loss, because of the extra I/O costs associated with the
> space used by all those padding blanks. I can think of very very few
> applications where CHAR(n) is really a sensible choice over VARCHAR(n).
> US state postal codes (CHAR(2)) are an example, but beyond that the
> valid use cases are mighty thin on the ground.
>
> Bruce, it occurs to me that the "Performance Tips" section of the user's
> guide ought to have an explanation of the performance implications of
> TOAST. We have some material in the FAQ but there's precious little
> mention in the manual. Thoughts?

Yes, I am working on my O'Reilly slides now and am dealing with how to
address TOAST for admins, specifically for disk space computations,
which I did cover in the new 'disk usage' section of the manual.

And in talking to the questioner, I was wondering if we have this
auto-out-of-line TOAST description anywhere in the docs. If not, we
need to add that.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026