[Starlink] The "reasons" that bufferbloat isn't a problem

[Gert Doering]

Hi,

On Fri, Jun 07, 2024 at 07:36:36AM +0200, Sebastian Moeller via Starlink wrote:
> it might show that big iron silicon is just inferior to general purpose CPUs

Well, different criteria for "inferior", I'd say.  A general purpose
CPU won't be able to feed something with 24x 100Gbit - and a switch
network processor won't be good for running a database...

For queueing/shaping, the network chips have become surprisingly good -
not sure about libreqos & friends, but "doing proper shaping to sub-
linerate to avoid policing drops in a carrier network further downstream"
was something switches just couldn't do, and recent gear (Jericho 2+)
does this quite well.  With proper burst size configured, no bufferbloat
there either :)

Gert Doering
        -- NetMaster
-- 
have you enabled IPv6 on something today...?

SpaceNet AG                      Vorstand: Sebastian v. Bomhard, Ingo Lalla,
                                           Karin Schuler, Sebastian Cler
Joseph-Dollinger-Bogen 14        Aufsichtsratsvors.: A. Grundner-Culemann
D-80807 Muenchen                 HRB: 136055 (AG Muenchen)
Tel: +49 (0)89/32356-444         USt-IdNr.: DE813185279