| From: | Peter Geoghegan <pg(at)bowt(dot)ie> |
|---|---|
| To: | "Campbell, Lance" <lance(at)illinois(dot)edu> |
| Cc: | "pgsql-admin(at)lists(dot)postgresql(dot)org" <pgsql-admin(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: Temporary Files |
| Date: | 2021-03-29 20:49:09 |
| Message-ID: | CAH2-Wz=F7-92QM7nYE_-JrjUhPVbHo4npdDYQZruFdwy+JYO_A@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-admin |
On Mon, Mar 29, 2021 at 11:10 AM Campbell, Lance <lance(at)illinois(dot)edu> wrote:
> If you divide 451 by 21 you get 21.4 . I would have expected to see an average size greater than 24 MB. This implies to me that PostgreSQL is creating temp files that are smaller than 24 MB.
>
> What am I not understanding?
work_mem limits the in-memory space used by sort and hash operations.
It doesn't necessarily follow that the on-disk footprint is exactly
(or even approximately) the same as the in-memory footprint for any
given operation that barely can't fit in memory. The precise space
overheads will vary based on lots of factors, most of which are
implementation details.
--
Peter Geoghegan
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Ivan Petrov | 2021-03-29 23:20:30 | Improve postgres 13.2 performance for concurrent bulk insert |
| Previous Message | Campbell, Lance | 2021-03-29 18:10:39 | Temporary Files |