One last point - much simpler and less cpu expensive than shaping is having this stuff on by default on all interfaces at line rate. This is the case now for OSX, iOS, and most linuxes today. Any place where you have a fast → slow transition some form of these new AQM and/or FQ technologies help, be it 10Gbits into 1Gbit, multiple ports into one port, etc. For more info, please see: RFC 7567 - IETF Recommendations Regarding Active Queue Management
1 Like