This message was deleted.

It is possible that the bandwidth estimation gives less than 1.5 Mbps. You would have to look at the logs for the actual value. There could be audio, signalling and potentially other traffic on the same down stream link too. So, the estimate may be less than 1.5 Mbps

OK, I'm currently neither using audio nor data. I created another profile, which throttled to 2.5 MBit/s. This time it worked as expected. Is there any "higher level" logging server-side for the BWE? Or some clever "grep" to be applicable to the logs to just filter the BWE relevant loggings?

grep for

stream allocator

Yepp, did that already. Still a bit overwhelming...

This is the most relevant line.

Which line do you mean?

That log line in the link ☝🏽

The link shows me the entire module code

Line #828 is what that link is pointing to. The log in that line is logged when congestion is detect and needs to be acted upon.

Well, a lot of cryptic stuff 🙂 2023-12-07T134231.819Z INFO livekit.transport.cc streamallocator/streamallocator.go:828 stream allocator: channel congestion detected, applying channel capacity update {"room": "abcde", "roomID": "RM_iZyTWaVuYmQi", "participant": "LKTest", "pID": "PA_WM459azCuuZu", "remote": false, "transport": "SUBSCRIBER", "reason": "LOSS", "old(bps)": 1483920, "new(bps)": 464078, "lastReceived(bps)": 464078, "expectedUsage(bps)": 1495225, "commitThreshold(bps)": 1420463, "channel": "name: non-probe, estimate: {n: non-probe-estimate, t: Thu Dec 7 134225 UTC 2023|Thu Dec 7 134231 UTC 2023|6.26s, v: 107|461440|464078|[461440(0), 464078(193), 464078(780), 464078(1380), 464078(1895), 464078(2396)]|1.00}, nack {n: non-probe-nack, t: Thu Dec 7 134229 UTC 2023|Thu Dec 7 134231 UTC 2023|2.01s, p: 376, rn: 207, rn/p: 0.55}"} 2023-12-07T134231.820Z INFO livekit.transport.cc streamallocator/streamallocator.go:838 stream allocator: channel congestion detected, applying channel capacity: experimental {"room": "abcde", "roomID": "RM_iZyTWaVuYmQi", "participant": "LKTest", "pID": "PA_WM459azCuuZu", "remote": false, "transport": "SUBSCRIBER", "rateHistory": ["t: 1701956551697, e: 463274.24, m: 1497373.46/2181835.29, um: 0.00/0.00, qd: 6.93", "t: 1701956551698, e: 463754.28, m: 1495503.68/2230149.57, um: 0.00/0.00, qd: 7.03", "t: 1701956551708, e: 451555.60, m: 1474036.50/2318107.32, um: 0.00/0.00, qd: 7.20", "t: 1701956551723, e: 458846.37, m: 1511290.88/2310399.85, um: 0.00/0.00, qd: 7.25", "t: 1701956551750, e: 446699.55, m: 1453692.36/2321939.82, um: 0.00/0.00, qd: 7.17", "t: 1701956551756, e: 449719.06, m: 1483451.44/2373421.59, um: 0.00/0.00, qd: 7.34", "t: 1701956551765, e: 453788.15, m: 1470140.06/2530730.18, um: 0.00/0.00, qd: 7.64", "t: 1701956551772, e: 457036.27, m: 1459700.16/2512758.72, um: 0.00/0.00, qd: 7.58", "t: 1701956551805, e: 454227.29, m: 1471553.84/2500537.24, um: 0.00/0.00, qd: 7.58", "t: 1701956551819, e: 461077.21, m: 1505647.41/2504787.55, um: 0.00/0.00, qd: 7.65"], "expectedQueuing": 4.222080445994146, "nackHistory": [], "trackHistory": {"TR_VCAiqhSyFghWDR":"t: 2023-12-07 134231.820054186 +0000 UTC m=+7466.084789126, n: [t: 1701956550865, l: 47654, h: 47684, sp: 31, nnd: 19, dens: 0.61, nns: 31, int: 1.63, nr: 7 t: 1701956550965, l: 47654, h: 47703, sp: 50, nnd: 14, dens: 0.28, nns: 21, int: 1.50, nr: 5 t: 1701956551065, l: 47659, h: 47717, sp: 59, nnd: 29, dens: 0.49, nns: 62, int: 2.14, nr: 12 t: 1701956551164, l: 47679, h: 47744, sp: 66, nnd: 40, dens: 0.61, nns: 75, int: 1.88, nr: 17 t: 1701956551265, l: 47701, h: 47751, sp: 51, nnd: 21, dens: 0.41, nns: 46, int: 2.19, nr: 8 t: 1701956551364, l: 47714, h: 47778, sp: 65, nnd: 37, dens: 0.57, nns: 73, int: 1.97, nr: 16 t: 1701956551464, l: 47736, h: 47792, sp: 57, nnd: 25, dens: 0.44, nns: 51, int: 2.04, nr: 10 t: 1701956551564, l: 47755, h: 47809, sp: 55, nnd: 28, dens: 0.51, nns: 58, int: 2.07, nr: 11 t: 1701956551665, l: 47774, h: 47826, sp: 53, nnd: 35, dens: 0.66, nns: 62, int: 1.77, nr: 14 t: 1701956551764, l: 47788, h: 47845, sp: 58, nnd: 44, dens: 0.76, nns: 88, int: 2.00, nr: 18], rr: [t: Thu Dec 7 134229 UTC 2023, l: 0, p: 8, rtt: 0 t: Thu Dec 7 134230 UTC 2023, l: 4294967289, p: 17, rtt: 0 t: Thu Dec 7 134230 UTC 2023, l: 0, p: 18, rtt: 0 t: Thu Dec 7 134230 UTC 2023, l: 0, p: 11, rtt: 0 t: Thu Dec 7 134230 UTC 2023, l: 11, p: 29, rtt: 0 t: Thu Dec 7 134230 UTC 2023, l: 15, p: 32, rtt: 0 t: Thu Dec 7 134231 UTC 2023, l: 24, p: 41, rtt: 0 t: Thu Dec 7 134231 UTC 2023, l: 16, p: 33, rtt: 0 t: Thu Dec 7 134231 UTC 2023, l: 19, p: 37, rtt: 0 t: Thu Dec 7 134231 UTC 2023, l: 24, p: 36, rtt: 0]"}}

This is the most relevant

"old(bps)": 1483920, "new(bps)": 464078, "lastReceived(bps)": 464078, "expectedUsage(bps)": 1495225,

It is expecting to use 1.49 Mbps (probably it was at mid layer), previously it had an estimate of 1.483 Mbps, but the new estimate is 464 kbps. So, it would drop down to the lowest layer.

At the moment I'm checking goalBps and probeGoalDeltaBps. And there is some margin (understandably): https://github.com/livekit/livekit/blob/b0b7489b0e1be36f035655826d2b52736ee020b3/pkg/sfu/streamallocator/probe_controller.go#L208