On 2014-04-28 10:03:58 -0400, Tom Lane wrote:
> What I find much more worrisome about Andres' proposals is that he
> seems to be thinking that there are *no* other changes to the buffer
> headers on the horizon.
Err. I am not thinking that at all. I am pretty sure I never made that
argument. The reason I want to limit the number of connections is it
allows *both*, shrinking the size of BufferDescs due to less alignment
padding *and* stuffing the refcount and flags into one integer.
> That's untenable. And I don't want to be told that we can't improve
> the buffer management algorithms because adding another field would
> make the headers not fit in a cacheline.
I think we need to move some less frequently fields to a separate array
to be future proof. Heikki suggested freeNext, wait_backend_pid I added
io_in_progress_lock. We could theoretically replace buf_id by
calculating it based on the BufferDescriptors array, but that's probably
not a good idea.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services
