excerpts from simultaneous netperf (AP -> wifi clients) using nmba and "59" client (the one that did 0.07 mbps BE upload in the tests above). Both clients in the same location as the test above. MCS values as reported on the AP (I checked several times during the ~5 min netperfs - they did not change).
59:
Interim result: 90.16 10^6bits/s over 1.001 seconds ending at 1656515800.668
Interim result: 88.43 10^6bits/s over 1.019 seconds ending at 1656515801.687
Interim result: 87.33 10^6bits/s over 1.013 seconds ending at 1656515802.700
Interim result: 88.50 10^6bits/s over 1.000 seconds ending at 1656515803.700
Interim result: 89.57 10^6bits/s over 1.005 seconds ending at 1656515804.705
Interim result: 88.42 10^6bits/s over 1.013 seconds ending at 1656515805.717
Interim result: 89.54 10^6bits/s over 1.005 seconds ending at 1656515806.722
Interim result: 89.47 10^6bits/s over 1.001 seconds ending at 1656515807.723
Interim result: 88.54 10^6bits/s over 1.013 seconds ending at 1656515808.736
Interim result: 87.97 10^6bits/s over 1.008 seconds ending at 1656515809.744
Interim result: 89.95 10^6bits/s over 1.003 seconds ending at 1656515810.747
nmba:
Interim result: 13.06 10^6bits/s over 2.077 seconds ending at 1656515800.050
Interim result: 13.27 10^6bits/s over 1.002 seconds ending at 1656515801.053
Interim result: 18.14 10^6bits/s over 1.002 seconds ending at 1656515802.054
Interim result: 16.41 10^6bits/s over 1.100 seconds ending at 1656515803.154
Interim result: 14.09 10^6bits/s over 1.163 seconds ending at 1656515804.317
Interim result: 14.41 10^6bits/s over 1.015 seconds ending at 1656515805.332
Interim result: 13.51 10^6bits/s over 1.066 seconds ending at 1656515806.398
Interim result: 12.35 10^6bits/s over 1.090 seconds ending at 1656515807.488
Interim result: 14.68 10^6bits/s over 1.015 seconds ending at 1656515808.503
Interim result: 19.06 10^6bits/s over 1.013 seconds ending at 1656515809.516
Interim result: 17.03 10^6bits/s over 1.121 seconds ending at 1656515810.637
I did not make a cut and paste error nor did I miss lablel the netperf results.
MCS from r7500v2 AP
r7500v2 # iw dev wlan0 station dump | grep "tx bitrate"
nmba tx bitrate: 360.0 MBit/s VHT-MCS 8 40MHz short GI VHT-NSS 2
59 tx bitrate: 180.0 MBit/s MCS 12 40MHz short GI
Both 59 and nmba are about 10 m from the AP, no clear line of site, and about 3 m from each other. There are gaps in the (wood frame and drywall) walls obstructing their line of site.
I started the 59 wifi client netperf first (this use to matter in the past - atm i can start either client first and get reproducible results). A single client netperf to nmba at this location can achieve 300+ mbps. When I start the nmba netperf fisrt, I'll see 200-300+ mbps until I start the 59 netperf at which point it drops to 10-20 mbps.
I'll bet that if I put tbf's on the wifi clients and limit their throughput to about 30 mbps and then repeat the "reversed" flent rtt_fair test, I'll get more meaningful results.
I wish the pre-packaged netperf binaries had the "-w" and "-b" options included by default. It would make using flent for this kind of testing a lot easier than playing with qdisc's.
EDIT 0: I added the 56 wifi client back in and tried simultaneous netperf's. 56 also is in the same location are the original reverse rtt_fair test above (~1 m from AP, clear line of site)
In words: I started the 56 client netperf first, then added nmba and 59. Throughput on the 56 client went from ~165 mbps (before starting nmba and 59) to a complete stop.
MCS 56 client stream on it's own:
r7500v2 # iw dev wlan0 station dump | grep "tx bitrate"
nmba tx bitrate: 300.0 MBit/s VHT-MCS 7 40MHz short GI VHT-NSS 2
59 tx bitrate: 180.0 MBit/s MCS 12 40MHz short GI
56 tx bitrate: 270.0 MBit/s MCS 15 40MHz
MCS all three clients streaming:
r7500v2 # iw dev wlan0 station dump | grep "tx bitrate"
nmba tx bitrate: 360.0 MBit/s VHT-MCS 8 40MHz short GI VHT-NSS 2
59 tx bitrate: 150.0 MBit/s MCS 7 40MHz short GI
56 tx bitrate: 30.0 MBit/s MCS 8 40MHz short GI
So something like this probably did happen during the reverse rtt_fair test. The only way I've been able to avoid it is to limit the total throughput to something the AP can handle (about 100 mbps by my estimation with this configuration).
ty for the suggestion to look at MCS.
EDIT 1: I turned the mac (nmba) off and added a third ubuntu wifi client (call it 135).
3 client (56, 59, & 135) simultaneous netperf:
56: ~30 mbps; tx bitrate: 150.0 MBit/s MCS 7
59: ~60 mbps; tx bitrate: 270.0 MBit/s MCS 14
135: ~70 mbps; tx bitrate: 300.0 MBit/s MCS 15
So ATF works for me? Or ATF works only if sans apple?