Feedback-ID: id4a34324:Fastmail
Date: Thu, 14 Aug 2025 16:44:14 -0400
From: Andres Freund <andres@anarazel.de>
To: Peter Geoghegan <pg@bowt.ie>
Cc: Tomas Vondra <tomas@vondra.me>, Thomas Munro <thomas.munro@gmail.com>,
	Nazir Bilal Yavuz <byavuz81@gmail.com>, Robert Haas <robertmhaas@gmail.com>,
	Melanie Plageman <melanieplageman@gmail.com>,
 PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
	Georgios <gkokolatos@protonmail.com>,
 Konstantin Knizhnik <knizhnik@garret.ru>,
	Dilip Kumar <dilipbalaut@gmail.com>
Subject: Re: index prefetching
Message-ID: <kvyser45imw3xmisfvpeoshisswazlzw35el3fq5zg73zblpql@f56enfj45nf7>
References: 
 <CAH2-Wz=L7h-koDKa3_NEg39Faw7MrOkSVOsodvQ4toSQahvWjQ@mail.gmail.com>
 <a9c0a5f9-cf5e-4ac9-abf6-683483161169@vondra.me>
 <DC1PBQW5HBEJ.3W4SMGI4LW3VI@bowt.ie>
 <7c308720-5cd4-4508-8cc0-6e55a6352731@vondra.me>
 <DC2B4TY023QP.2ZL331BPZCBSX@bowt.ie>
 <CAH2-WzmuGzTH-62EWTgQ4F66XEBJtJk25psF4GDuAGqeC4a34g@mail.gmail.com>
 <DC2DFSYXBDSJ.2N7WXLD4JT05H@bowt.ie>
 <6wyxbnry2unm3kbcu2sabhzhs7baoedlg77xqm42chpofjq45g@igst42zpl7ok>
 <CAH2-WzntgDeopLJpyEbUh23Qr1vgoYv5jbFkYsymTScEKxBj7A@mail.gmail.com>
 <CAH2-WzkaTHg2X9R-gLRNBEoL82t2mkrQq-3f=y3GAzrj40fFZw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CAH2-WzkaTHg2X9R-gLRNBEoL82t2mkrQq-3f=y3GAzrj40fFZw@mail.gmail.com>
Archived-At: 
 <https://www.postgresql.org/message-id/kvyser45imw3xmisfvpeoshisswazlzw35el3fq5zg73zblpql%40f56enfj45nf7>
Precedence: bulk

Hi,

On 2025-08-14 15:45:26 -0400, Peter Geoghegan wrote:
> On Thu, Aug 14, 2025 at 3:15 PM Peter Geoghegan <pg@bowt.ie> wrote:
> > Then why does the exact same pair of runs show "I/O Timings: shared
> > read=194.629" for the sequential table backwards scan (with total
> > execution time 1132.360 ms), versus "I/O Timings: shared read=352.88"
> > (with total execution time 697.681 ms) for the random table backwards
> > scan?
>
> If you're interested in trying this out for yourself, I've pushed my
> working branch here:
>
> https://github.com/petergeoghegan/postgres/tree/index-prefetch-batch-v1.2
>
> Note that the test case you'll run is added by the most recent commit:
>
> https://github.com/petergeoghegan/postgres/commit/c9ceb765f3b138f53b7f1fdf494ba7c816082aa1
>
> Run microbenchmarks/random_backwards_weird.sql to do an initial load
> of both of the tables. Then run
> microbenchmarks/queries_random_backwards_weird.sql to actually run the
> relevant queries. There are 4 such queries, but only the 2 backwards
> scan queries really seem relevant.

Interesting. In the sequential case I see some waits that are not attributed
in explain, due to the waits happening within WaitIO(), not WaitReadBuffers().
Which indicates that the read stream is trying to re-read a buffer that
previously started being read.

   read_stream_start_pending_read()
-> StartReadBuffers()
-> AsyncReadBuffers()
-> ReadBuffersCanStartIO()
-> StartBufferIO()
-> WaitIO()

There are far fewer cases of this in the random case.