Re: Progress on fast path sorting, btree index creation time

Поиск
Список
Период
Сортировка
От Tom Lane
Тема Re: Progress on fast path sorting, btree index creation time
Дата
Msg-id 16818.1328712665@sss.pgh.pa.us
обсуждение исходный текст
Ответ на Re: Progress on fast path sorting, btree index creation time  (Peter Geoghegan <peter@2ndquadrant.com>)
Ответы Re: Progress on fast path sorting, btree index creation time  (Robert Haas <robertmhaas@gmail.com>)
Список pgsql-hackers
Peter Geoghegan <peter@2ndquadrant.com> writes:
> It doesn't necessarily matter if we increase the size of the postgres
> binary by 10%, precisely because most of that is not going to be in
> play from one instant to the next.

You've heard of swapping, no?  Basically what I'm hearing from you is
total denial that binary bloat costs anything, and that just does not
square with reality.  Even if the costs in performance terms are
negligible in many common situations (something that you've asserted
but without offering any proof), there are other costs; software
distribution for instance is definitely sensitive to size.  As a Red Hat
person I've had to spend time fighting to keep Postgres included in the
minimum "does it fit on a DVD" distribution of RHEL/Fedora.  That case
gets harder to make every year, and yet it's pretty critical to the
success of the project --- if you don't get distributed, you lose users.

IMO this patch is already well past the point of diminishing returns in
value-per-byte-added.  I'd like to see it trimmed back to provide a fast
path for just single-column int4/int8/float4/float8 sorts.  The other
cases aren't going to offer enough of a win to justify the code space.
        regards, tom lane


В списке pgsql-hackers по дате отправления:

Предыдущее
От: Alvaro Herrera
Дата:
Сообщение: Re: Vacuum rate limit in KBps
Следующее
От: Robert Haas
Дата:
Сообщение: Re: Vacuum rate limit in KBps