thank you for bbrv3!

[Christopher Zeimke]

1. I just want to share my enthusiasm for how awesome it is that a distribution like CachyOS already includes your work on BBRv3 via a kernel patch. It makes it incredibly easy for end-users like me to benefit from the latest version without having to patch the kernel myself.

2. Your work has a direct, tangible impact. On my Wi-Fi connection, large downloads under Linux were a pain with the default CUBIC algorithm—the speed fluctuated wildly and constantly collapsed. Switching to BBRv3 is what finally stabilized my connection, allowing me to consistently utilize my full bandwidth. The difference is like night and day.

THANK YOU FOR YOUR WORK! :)

[Neal Cardwell]

Thanks for your kind words!

Just double-checking on the details of your download tests: for your "large downloads", were you changing the congestion control algorithm of the server / data sender from CUBIC to BBRv3? Or changing the congestion control algorithm of the client / data receiver from CUBIC to BBRv3?

From this brief description it sounds a bit like you changed the congestion control algorithm on a client  device (connected via WiFi) from CUBIC to BBRv3. But that change would not be expected to change the performance of downloads to that device, since CUBIC and BBRv3 are sender-based congestion control algorithms, where the TCP endpoint of the data sender is making the congestion control decisions about the sending rate. So I wanted to double-check. :-)

Best regards,

neal

 

--
You received this message because you are subscribed to the Google Groups "BBR Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bbr-dev+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/bbr-dev/bc589236-221f-4b33-97e0-db7217ce632cn%40googlegroups.com.

[DR. MUHAMMAD AHSAN]

I guess he experienced a "placebo effect" after shifting to bbr v3 :)

Regards,
Dr. Ahsan

To view this discussion visit https://groups.google.com/d/msgid/bbr-dev/CADVnQynTU52xxaVWVkwnu6H7mXdW0Fy2pJPUUM5%3D7GKGuoMncQ%40mail.gmail.com.

[Christopher Zeimke]

Wow, thanks so much for the explanation!  My apologies for the incorrect assumption in my first post. I honestly had no idea it was a sender-side algorithm, that's a really interesting detail.

You were right to double-check! I looked at my notes again and completely forgot that I had changed something else at the same time. When I set the algorithm to BBR, I also set "net.core.default_qdisc = fq."

So I guess the huge, real-world difference I saw on my Wi-Fi wasn't BBR on my end, but the effect of fq? Maybe it smoothed out my connection so the acknowledgements my PC sent back were much more stable, which then helped the server's congestion control to keep the speed high.

Thanks again for your work and for teaching me something new!

Best regards, Christopher