On Apr 13, 2023, at 3:13 PM, Dave Taht via Rpm <rpm@lists.bufferbloat.net> wrote:
I have not been paying a lot of attention here of late. I objected to
the enormous number of flows goresponsiveness was doing, and strongly
suggested it run to time of first loss or mark.
I find 4 flows enough
to stress out a network.
4 flows should often be enough to utilize a network at its full capacity. However, are 4 flows going to provide you stable working conditions in such a way that the bottleneck’s buffers are entirely full over an extended period without fluctuations due to the congestion-response of the individual flows? Also, what is the convergence time to reach stable working conditions and full buffer utilization at 4 flows?
Now, that being said. I don’t think that the exact number of flows is that important. They just need to be enough to fill the buffers in a stable way. All nit-picking on the number of flows just distracts us from actually solving problems on the Internet.
Anyway, is the structure of networkQuality
changing any in ippm?
Discussions around the methodology and the development of the tools as well as addressing issues we find in open-source networking stacks is all happening at the “Network-quality community”, at
https://github.com/network-quality/community/wiki. We have a slack-channel and a weekly meeting. Everyone who is actively working on development is invited to join our slack-channel and attend the weekly meetings. (The "active development" is a strong requirement )
Cheers,
Christoph
Some context here:
https://www.reddit.com/r/amazoneero/comments/12ksu9d/sqm_optimizing_for_videoconferencing_and_gaming/jg4zsq2/?context=8&depth=9
--
AMA March 31: https://www.broadband.io/c/broadband-grant-events/dave-taht
Dave Täht CEO, TekLibre, LLC
_______________________________________________
Rpm mailing list
Rpm@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/rpm