Horton

Re: AIO / read stream heuristics adjustments for index prefetching

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Melanie Plageman <melanieplageman@gmail.com>
Cc: pgsql-hackers@postgresql.org, Thomas Munro <thomas.munro@gmail.com>, Peter Geoghegan <pg@bowt.ie>, Tomas Vondra <tv@fuzzy.cz>, Nazir Bilal Yavuz <byavuz81@gmail.com>
Date: 2026-04-03T20:41:18Z
Lists: pgsql-hackers 
Hi,

Forgot to send this one earlier:

On 2026-04-03 11:46:59 -0400, Melanie Plageman wrote:
> On Thu, Apr 2, 2026 at 11:47 AM Andres Freund <andres@anarazel.de> wrote:
> >
> > What do you think about the updated patch to achieve that that I posted?
> 
> here is some review on 0005 and 0006 earlier posted
> concrete things:
> -------
> - I’d reorder stream→distance == 0 in read_stream_look_ahead() (and
> issue), I found myself asking why it wasn’t first

It'd not be correct in should_issue_now(). We can't move it before

	/* there is no pending IO that could be issued */
	if (pending_read_nblocks == 0)
		return false;

because then we'd trigger a call to read_stream_start_pending_read(), without
there being a pending read.

We can't move it before

	/* never start more IOs than our cap */
	if (stream->ios_in_progress >= stream->max_ios)
		return false;

because that could lead us to exceeding max_ios.

Greetings,

Andres Freund

Commits

  1. aio: io_uring: Trigger async processing for large IOs

  2. read stream: Split decision about look ahead for AIO and combining

  3. read_stream: Move logic about IO combining & issuing to helpers

  4. read_stream: Only increase read-ahead distance when waiting for IO

  5. read_stream: Prevent distance from decaying too quickly

  6. read_stream: Issue IO synchronously while in fast path

  7. aio: io_uring: Allow IO methods to check if IO completed in the background

  8. bufmgr: Return whether WaitReadBuffers() needed to wait