On 05/23/2015 15:54, Simon Barber wrote:
For throughput to be unaffected with a single TCP-Reno stream the buffer must grow to contain half the BDP, since the window will be halved on loss. With TCP-CUBIC the buffer must contain 20% of the BDP. CoDEL will make it's first drop when the sojourn time remains > 5ms, so with TCP-Reno, if the RTT is > 10ms, then utilization will be hurt. With TCP-CUBIC if the RTT is > 25ms then utilization will be hurt. Is this correct?
I've seen something similar in physical testbed experiments. Although not exactly what you're querying, http://caia.swin.edu.au/reports/140630A/CAIA-TR-140630A.pdf is related -- Section III.B notes NewReno throughput over a codel/pie/fq_codel bottleneck is markedly worse @ 200ms RTT than @ 20ms RTT (all other things being equal). Section IV.B sees the same with CUBIC. cheers, gja _______________________________________________ aqm mailing list [email protected] https://www.ietf.org/mailman/listinfo/aqm
