On: End-to-end congestion control cannot avoid latency spikes (2022)

[dilyevsky]

There's a paper by Google where they demonstrated that you can successfully push utilization far beyond of what is suggested here[0]

[0] - https://www.cs.princeton.edu/courses/archive/fall17/cos561/p...

[kccqzy → dilyevsky]

Section 6.2 of the paper says Google does traffic engineering (i.e. designating some traffic to be latency tolerant). That's just what solution 3 is in the article. For a single company, of course it can prioritize different traffic differently. It would be more troublesome for an ISP to do the same.

[sambazi → dilyevsky]

looks like they only care for drops and not latency

[dilyevsky → sambazi]

Drops is what happens when the queue is filling up (via AQM) but they track queue utilization as well