tgl@sss.pgh.pa.us (Tom Lane) wrote:
> I'm willing to do the code changes to separate TOAST_THRESHOLD from
> the toast chunk size, but I do not have the time or facilities to do
> any performance testing for different parameter choices. Anyone
> want to work on that?
What have you got in mind there? I might be able to do something,
assuming that it doesn't cause heartburn that I'll be "offline" April
6-14th.
Are we simply talking about having the option of #defining a different
threshold at which items get thrown out to TOAST? Or one of the more
sophisticated options? Or is the idea more simply that we might
consider having the default set somewhat lower than it is at present?
And I guess a good question is, what's going to get regarded as a
meaningful test? I've got a couple local test cases I could draw
from, unfortunately, the interaction with TOAST will more than likely
be pretty trivial, showing off "Yeah, cutting the threshold was a good
idea." And that may not be fair to everyone's case.
[The good news is, of course, that if the end "deliverable" is a
single #define parameter that's used as the denominator to the
fraction, delivery during "beta" time is, indeed, quite trivial...]
--
output = ("cbbrowne" "@" "gmail.com")
http://linuxdatabases.info/info/slony.html
Outside of a dog, a book is man's best friend. Inside of a dog, it's
too dark to read. -Groucho Marx