Re: Postgresql network transmission overhead

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: "Michael J(dot) Baars" <mjbaars1977(dot)pgsql-hackers(at)cyberfiber(dot)eu>
Cc: pgsql-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: Postgresql network transmission overhead
Date: 2021-02-26 15:11:57
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

"Michael J. Baars" <mjbaars1977(dot)pgsql-hackers(at)cyberfiber(dot)eu> writes:
> In the logfile you can see that the effective user data being written is only 913kb, while the actual being transmitted over the network is 7946kb when writing
> one row at a time. That is an overhead of 770%!

So ... don't write one row at a time.

You haven't shown any details, but I imagine that most of the overhead
comes from per-query stuff like the RowDescription metadata. The intended
usage pattern for bulk operations is that there's only one RowDescription
message for a whole lot of data rows. There might be reasons you want to
work a row at a time, but if your concern is to minimize network traffic,
don't do that.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Joel Jacobson 2021-02-26 16:42:32 Re: Some regular-expression performance hacking
Previous Message 2021-02-26 14:53:58 RE: [HACKERS] logical decoding of two-phase transactions