| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> |
| Cc: | pgsql-hackers(at)postgresql(dot)org, mkellycs(at)gmail(dot)com, ashutosh(dot)bapat(at)enterprisedb(dot)com |
| Subject: | Re: [POC] FETCH limited by bytes. |
| Date: | 2015-01-22 16:17:52 |
| Message-ID: | 24503.1421943472@sss.pgh.pa.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> writes:
> Hello, as the discuttion on async fetching on postgres_fdw, FETCH
> with data-size limitation would be useful to get memory usage
> stability of postgres_fdw.
> Is such a feature and syntax could be allowed to be added?
This seems like a lot of work, and frankly an incredibly ugly API,
for a benefit that is entirely hypothetical. Have you got numbers
showing any actual performance win for postgres_fdw?
Even if we wanted to do something like this, I strongly object to
measuring size by heap_compute_data_size. That's not a number that users
would normally have any direct knowledge of; nor does it have anything
at all to do with the claimed use-case, where what you'd really need to
measure is bytes transmitted down the wire. (The difference is not small:
for instance, toasted values would likely still be toasted at the point
where you're measuring.)
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tomas Vondra | 2015-01-22 16:18:19 | Re: PATCH: decreasing memory needlessly consumed by array_agg |
| Previous Message | Petr Jelinek | 2015-01-22 16:02:36 | Re: Sequence Access Method WIP |