public inbox for passt-dev@passt.top
 help / color / mirror / code / Atom feed
From: Max Chernoff <git@maxchernoff.ca>
To: Stefano Brivio <sbrivio@redhat.com>
Cc: passt-dev@passt.top, David Gibson <david@gibson.dropbear.id.au>
Subject: Re: [PATCH v3 00/10] tcp: Fix throughput issues with non-local peers
Date: Mon, 08 Dec 2025 01:51:48 -0700	[thread overview]
Message-ID: <6c87ae32518893269fa4a99c120f3990972242f9.camel@maxchernoff.ca> (raw)
In-Reply-To: <20251208092530.2638636a@elisabeth>

Hi Stefano,

On Mon, 2025-12-08 at 09:25 +0100, Stefano Brivio wrote:
> but, in general, I don't think things can work reliably
> with the values you shared for tcp_notsent_lowat.

Ok, that works for me. I know very little about TCP, so I just blindly
copied that value for tcp_notsent_lowat from

    https://blog.cloudflare.com/http-2-prioritization-with-nginx/

but if that's incompatible with pasta, then I have no problem resetting
tcp_notsent_lowat back to the kernel default.

A random web search makes it look like changing tcp_notsent_lowat is
somewhat common

    https://www.google.com/search?q=tcp_notsent_lowat%3D131072

    https://github.com/search?q=tcp_notsent_lowat%3D131072+NOT+is%3Afork&type=code

so maybe it would be a good idea for pasta to either use setsockopt to
override it, or to print a warning on startup if the sysctl is set too
low?

> Does this (upload now taking longer/timing out with 50 ms RTT) also
> happen without "custom" values for tcp_notsent_lowat?
>
> I tested things quite extensively in that RTT region (without custom
> sysctl values) and the improvement looks rather consistent to me.

Ok, with tcp_notsent_lowat reset to the Fedora defaults, the upload
speeds with large RTTs do indeed look *much* better

    $ sudo dnf install <freshly built pasta rpms>
    $ pasta --version
    pasta 0^20251208.g5943ea4-1.fc43.x86_64

    $ ./pasta-upload-test.sh
    network ping_time       wmem_max        rmem_max        tcp_notsent_lowat       tcp_congestion_control  default_qdisc        download_time   upload_time
    host    50ms    custom  custom  default  custom  custom  1.561761        2.045501
    pasta   50ms    custom  custom  default  custom  custom  1.575290        1.707500
    host    170ms   custom  custom  default  custom  custom  9.147689        3.220591
    pasta   170ms   custom  custom  default  custom  custom  13.351799       3.411078

    $ sudo dnf downgrade pasta
    $ pasta --version  # Back to the stock pasta in F43
    pasta 0^20250919.g623dbf6-1.fc43.x86_64

    $ ./pasta-upload-test.sh
    network ping_time       wmem_max        rmem_max        tcp_notsent_lowat       tcp_congestion_control  default_qdisc        download_time   upload_time
    host    50ms    custom  custom  default  custom  custom  1.429540        1.674165
    pasta   50ms    custom  custom  default  custom  custom  1.503907        2.025471
    host    170ms   custom  custom  default  custom  custom  8.891267        3.039416
    pasta   170ms   custom  custom  default  custom  custom  11.056843       18.704653

Thanks again for all your help,
-- Max

  reply	other threads:[~2025-12-08  8:51 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-08  7:20 Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 01/10] tcp, util: Add function for scaling to linearly interpolated factor, use it Stefano Brivio
2025-12-09  5:05   ` David Gibson
2025-12-08  7:20 ` [PATCH v3 02/10] tcp: Change usage factor of sending buffer in tcp_get_sndbuf() to 75% Stefano Brivio
2025-12-09  5:05   ` David Gibson
2025-12-08  7:20 ` [PATCH v3 03/10] tcp: Limit advertised window to available, not total sending buffer size Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 04/10] tcp: Adaptive interval based on RTT for socket-side acknowledgement checks Stefano Brivio
2025-12-09  5:10   ` David Gibson
2025-12-09 22:49     ` Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 05/10] tcp: Don't clear ACK_TO_TAP_DUE if we're advertising a zero-sized window Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 06/10] tcp: Acknowledge everything if it looks like bulk traffic, not interactive Stefano Brivio
2025-12-09  5:12   ` David Gibson
2025-12-08  7:20 ` [PATCH v3 07/10] tcp: Don't limit window to less-than-MSS values, use zero instead Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 08/10] tcp: Allow exceeding the available sending buffer size in window advertisements Stefano Brivio
2025-12-08  8:14   ` Max Chernoff
2025-12-08  8:15   ` Max Chernoff
2025-12-08  8:27     ` Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 09/10] tcp: Send a duplicate ACK also on complete sendmsg() failure Stefano Brivio
2025-12-08  7:20 ` [PATCH v3 10/10] tcp: Skip redundant ACK on partial " Stefano Brivio
2025-12-08  8:11 ` [PATCH v3 00/10] tcp: Fix throughput issues with non-local peers Max Chernoff
2025-12-08  8:25   ` Stefano Brivio
2025-12-08  8:51     ` Max Chernoff [this message]
2025-12-08  9:00       ` Stefano Brivio

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=6c87ae32518893269fa4a99c120f3990972242f9.camel@maxchernoff.ca \
    --to=git@maxchernoff.ca \
    --cc=david@gibson.dropbear.id.au \
    --cc=passt-dev@passt.top \
    --cc=sbrivio@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://passt.top/passt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).