My position remains that fq in the network opens up all possibilities for all forms of e2e congestion control.
Tests with BBR vs cubic and other tcps on all the single queued aqm designs I've tried (RED, codel, pie) have been pretty dismal. Tests against fq'd with 1024 on cake and fq_codel: nary a glitch. It would be nice if BBR had a sane response of any sort against ECN marking. I now have a ton of data on bbr vs everything i have (is there a patch for l4s that uses ECT(1)?), please let me know what sort of tests you would like me to run, or what out of my datasets are most worth writing up, as we are in the process of analyzing wifi behaviors post https://lwn.net/Articles/705884/ _______________________________________________ aqm mailing list [email protected] https://www.ietf.org/mailman/listinfo/aqm
