esnet
extremely high throughput
I run iperf with option "-u -b 10M"(ex1) and "-u -b 50M"(ex2), but at some point the bitrate output looks strange.
ex1) 15:20:34 --> 634Mbps >> 10M !!
(part of the output as below)
Tue Jul 1 15:20:07 2025 [ 5] 1926.00-1927.00 sec 1.22 MBytes 10.3 Mbits/sec 0.800 ms 0/886 (0%)
Tue Jul 1 15:20:08 2025 [ 5] 1927.00-1928.00 sec 1.19 MBytes 10.0 Mbits/sec 0.584 ms 0/865 (0%)
Tue Jul 1 15:20:09 2025 [ 5] 1928.00-1929.00 sec 1.17 MBytes 9.80 Mbits/sec 0.439 ms 0/846 (0%)
Tue Jul 1 15:20:10 2025 [ 5] 1929.00-1930.00 sec 1.21 MBytes 10.1 Mbits/sec 1.291 ms 0/875 (0%)
Tue Jul 1 15:20:11 2025 [ 5] 1930.00-1931.00 sec 973 KBytes 7.97 Mbits/sec 1.055 ms 35/723 (4.8%)
Tue Jul 1 15:20:12 2025 [ 5] 1931.00-1932.06 sec 467 KBytes 3.60 Mbits/sec 1.229 ms 0/330 (0%)
Tue Jul 1 15:20:13 2025 [ 5] 1932.06-1933.14 sec 472 KBytes 3.60 Mbits/sec 0.792 ms 279/613 (46%)
Tue Jul 1 15:20:14 2025 [ 5] 1933.14-1934.08 sec 560 KBytes 4.84 Mbits/sec 1.202 ms 445/841 (53%)
Tue Jul 1 15:20:34 2025 [ 5] 1934.08-1954.09 sec 781 KBytes 320 Kbits/sec 1.320 ms 353/905 (39%)
Tue Jul 1 15:20:34 2025 [ 5] 1954.09-1954.10 sec 139 KBytes 683 Mbits/sec 3.503 ms 0/98 (0%)
Tue Jul 1 15:20:35 2025 [ 5] 1954.10-1955.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
Tue Jul 1 15:20:36 2025 [ 5] 1955.00-1956.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
Tue Jul 1 15:20:37 2025 [ 5] 1956.00-1957.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
Tue Jul 1 15:20:38 2025 [ 5] 1957.00-1958.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
Tue Jul 1 15:20:39 2025 [ 5] 1958.00-1959.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
ex2) 08:21:17 --> 112Mbps >> 50M !!
(part of the output as below)
Thu Jul 17 08:21:13 2025 [ 5] 63473.00-63474.00 sec 1.79 MBytes 15.0 Mbits/sec 0.562 ms 3022/4320 (70%)
Thu Jul 17 08:21:14 2025 [ 5] 63474.00-63475.00 sec 1.69 MBytes 14.2 Mbits/sec 0.383 ms 3098/4320 (72%)
Thu Jul 17 08:21:15 2025 [ 5] 63475.00-63476.00 sec 897 KBytes 7.34 Mbits/sec 0.219 ms 1853/2487 (75%)
Thu Jul 17 08:21:16 2025 [ 5] 63476.00-63477.00 sec 998 KBytes 8.18 Mbits/sec 0.259 ms 36/742 (4.9%)
Thu Jul 17 08:21:17 2025 [ 5] 63477.00-63478.00 sec 13.4 MBytes 112 Mbits/sec 0.385 ms 541/10240 (5.3%)
Thu Jul 17 08:21:18 2025 [ 5] 63478.00-63479.00 sec 5.86 MBytes 49.2 Mbits/sec 0.299 ms 206/4451 (4.6%)
Thu Jul 17 08:21:19 2025 [ 5] 63479.00-63480.00 sec 5.69 MBytes 47.7 Mbits/sec 0.329 ms 199/4317 (4.6%)
Thu Jul 17 08:21:20 2025 [ 5] 63480.00-63481.00 sec 5.68 MBytes 47.6 Mbits/sec 0.278 ms 199/4313 (4.6%)
Thu Jul 17 08:21:21 2025 [ 5] 63481.00-63482.00 sec 5.69 MBytes 47.7 Mbits/sec 0.294 ms 199/4316 (4.6%)
Thu Jul 17 08:21:22 2025 [ 5] 63482.00-63483.00 sec 5.68 MBytes 47.7 Mbits/sec 0.290 ms 199/4315 (4.6%)
Anyone could tell me how this happened ?
For the "-u -b 50M"(ex2), it seems that the messages were buffered somewhere, probably in the server. This is as in the period before the 112Mbis/sec was relatively low - around 1MB/sec was received instead of around 5.7/sec. I.e., each second about 4-5MB were buffered. That can explain the 13.4MB that were received. This may be because the server CPU/etc. was loaded with other tasks. Then when the server became free it handled the buffered messages.
For the "-u -b 10M"(ex1) I don't have an explanation, as 139 KBytes are not 683 Mbits. Seems like a calculation error. Additional information may help the evaluation - can you send the output of iperf3 -v for both the client and server? Also, can you send the full client and server commands (no need for IP addresses)?
Regarding the "-u -b 10M"(ex1) test. I now notice that the last period reported with data is only 10ms (1954.09-1954.10). Probably at that period the iperf3 just read buffered messages and this is the reason for the "high throughput". I assume that this was at the end of the test, but for some reason the server didn't stop. Again, sending the iperf3 versions used and the commands used will help better understanding.
Thank you for your comment !! The exact command is as below(ex2 used tcp) ex1)iperf3 -c SERVER_IP -u -i 1 -p 5201 -t 86400 -b 10M --timestamps ex2)iperf3 -c SERVER_IP -i 1 -p 5201 --cport 53001 -t 86400 -b 50M -R --timestamps
Even if the 139KB data transfered for 100ms, bitrate seems to be too high to me. The buffer data might have caused calclulation error.
Even if the 139KB data transfered for 100ms, bitrate seems to be too high to me. The buffer data might have caused calclulation error.
Actually, it is 10ms not 100ms (my mistake), so maybe the calculation is not build for such short time.
In any case, the main issue in ex1 is why no more data was received past this point. Can you send the output of iperf3 -v for both the client and server? Also, did you see if the client continued to send data at this point, or maybe the problem was in the client?
Tue Jul 1 15:20:14 2025 [ 5] 1933.14-1934.08 sec 560 KBytes 4.84 Mbits/sec 1.202 ms 445/841 (53%) Tue Jul 1 15:20:34 2025 [ 5] 1934.08-1954.09 sec 781 KBytes 320 Kbits/sec 1.320 ms 353/905 (39%) Tue Jul 1 15:20:34 2025 [ 5] 1954.09-1954.10 sec 139 KBytes 683 Mbits/sec 3.503 ms 0/98 (0%) Tue Jul 1 15:20:35 2025 [ 5] 1954.10-1955.00 sec 0.00 Bytes 0.00 bits/sec 3.503 ms 0/0 (0%)
I just noticed that the reported interval before the extremely high throughput was 20 seconds. Any idea what may have caused this? I.e. system was loaded? paused? ... I run similar test on my computer and paused the server process for 20 seconds, then changed the system clock by 20 seconds, but none caused any issue.