re: https://arxiv.org/pdf/1905.07152.pdf
A nit with this paper is that I'd have preferred they measured and reported the actual rtts to the CDNs they tested, rather than inflate them 50ms via netem. They report all sorts of different behaviors on competing flows vs a vs akamai, cloudflare, edgecast, fastly, and google - big buffers, small, bbr vs cubic, etc. And, ok, ok, the final bit here after that eval happened is why I'm reposting widely. "Luckily, the flow-queuing variant of fq_codel enables a large degree of fairness even in heterogeneous settings. Thus, it seems to again stand that the technologies to enable a fair and performant Internet are available and only need to be deployed at the bottlenecks." Their prior work, categorizing IW sizes and behaviors in the real world, was also quite good. https://tma.ifip.org/2018/wp-content/uploads/sites/3/2018/06/tma2018_paper13.pdf -- Dave Täht CTO, TekLibre, LLC http://www.teklibre.com Tel: 1-831-205-9740 _______________________________________________ Bloat mailing list [email protected] https://lists.bufferbloat.net/listinfo/bloat
