I have been writing down some requirements for a new version of the CAKE qdisc based on observed results from the field.
I apologize for the somewhat random nature of this (and there are a great deal of outstanding comments on it - we seem to be seeing some issues with steam and ecn), I tend to write first and polish later. I started putting in some observations about starlink´s behaviors and how to possibly fix them, towards the end, in thinking about adding a "plug" facility to it in particular. https://docs.google.com/document/d/1tTYBPeaRdCO9AGTGQCpoiuLORQzN_bG3TAkEolJPh28/edit Anyway, if you don´t feel like wading through that mess: In looking over my starlink data accumulated in the last couple weeks I am seeing that 70ms spike more regularly, and I am curious if it is appearing in other data? Why are the steps in baseline latency are so large? Could they be doing something further to delay the packets and deal with the doppler that way? I keep hoping that a good video conferencing analysis tool would show up that would show glitches per minute and the recovery period for each glitch. -- Oct 30: https://netdevconf.info/0x17/news/the-maestro-and-the-music-bof.html Dave Täht CSO, LibreQos _______________________________________________ Starlink mailing list [email protected] https://lists.bufferbloat.net/listinfo/starlink
