From: | "Peter J(dot) Holzer" <hjp-pgsql(at)hjp(dot)at> |
---|---|
To: | pgsql-general(at)lists(dot)postgresql(dot)org |
Subject: | Re: COPY threads |
Date: | 2018-10-11 20:02:10 |
Message-ID: | 20181011200210.ldvty74xod7qw4zf@hjp.at |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
On 2018-10-10 17:19:50 -0400, Ravi Krishna wrote:
> > On Oct 10, 2018, at 17:18 , Andres Freund <andres(at)anarazel(dot)de> wrote:
> > On October 10, 2018 2:15:19 PM PDT, Ravi Krishna <srkrishna1(at)aol(dot)com> wrote:
> >> If I have a large file with say 400 million rows, can I first split it
> >> into 10 files of 40 million rows each and then fire up 10 different
> >> COPY sessions , each reading from a split file, but copying into the
> >> same table. I thought not. It will be great if we can do this.
> >
> > Yes, you can.
> >
> Thank you. Let me test it and see the benefit. We have a use case for this.
You should of course test this on your own hardware with your own data,
but here are the results of a simple benchmark (import 1 million rows
into a table without indexes via different methods) I ran a few weeks
ago on one of our servers:
y axis is rows per second. x axis are different runs, sorted from
slowest to fastest (so 2 is the median).
As you can see it doesn't parallelize perfectly: 2 copy processes are
only about 50 % faster than 1, and 4 are about 33 % faster than 2. But
there is a still quite a respectable performance boost.
hp
PS: The script is of course in the same repo, but I didn't include the
test data because I don't think I'm allowed to include that.
--
_ | Peter J. Holzer | we build much bigger, better disasters now
|_|_) | | because we have much more sophisticated
| | | hjp(at)hjp(dot)at | management tools.
__/ | http://www.hjp.at/ | -- Ross Anderson <https://www.edge.org/>
From | Date | Subject | |
---|---|---|---|
Next Message | Dmitry O Litvintsev | 2018-10-11 20:17:50 | something weird happened - can select by column value although column value exist |
Previous Message | Laurenz Albe | 2018-10-11 17:53:17 | Re: how to identify the timeline of specified recovery_target_time when do multiple PITR |