What does the value "bad_loss" in the output mean?

[trinhxhai2000]

I did the test below and got a high value of bad loss. With --video option, it gets worse even more 18092.21 %.

My questions are:

• What does the value "bad loss" in the output mean?

• How can the value be more than 100%?

• How can I improve the "bad loss" value?

Thank you very much !

Here is my test's output:

./turnhammer {my server ip}:3478 test test123 --audio -j 200 -d 60 --force

The test would do approx 11.878 Mbit/s and consume 79.189 megabytes of traffic
My external address: 42.113.144.74:52419
Allocated 200 TURN clients
Received the first datagram
Received 190682 packets from 192000 window of total 192000 || Loss: 00.6865%   bad loss: 49.6276%
RTT4 brackets: 0-49ms: 00.0000%   180-399ms: 09.8271%  1000-1999ms: 00.0828%
             50-179ms: 87.1000%   400-999ms: 02.3036%      2000+ms: 00.0000%
 <<<  Overall score:  -39.9 / 10.0  >>>
Stopping TURN clients

[vi]

What does the value "bad loss" in the output mean?

It is loss of 50 packets in a row. It is miscalculated in this case - the percentage should be between 0% and loss%. Maybe the algorithm gets confused by reordered packets or was not tested with -j.

How can the value be more than 100%?

Because of a bug in turnhammer.

How can I improve the "bad loss" value?

"Bad loss" calculator needs to be improved first. "Overall score" also gets affected by this.

---

Also note that CPU load on turnhammer side and on coturn side is similar - you may get poor results because of overloaded client, not only server. For a reliable load-test of a TURN server, you need to start turnhammer instances on multiple nodes simultaneously, pointed at one server.

[trinhxhai2000]

Thanks for your help.

[vi]

Tried running similar command line against my TURN server and it happens to work properly.

$ turnhammer ....:3478 ...  ... --audio -j 200 -d 60 --force
The test would do approx 11.878 Mbit/s and consume 79.189 megabytes of traffic
My external address: 46.53.246.99:6990
Allocated 200 TURN clients
Received the first datagram
Received 191555 packets from 192000 window of total 192000 || Loss: 00.2318%   bad loss: 00.0000%
RTT4 brackets: 0-49ms: 00.0000%   180-399ms: 99.7682%  1000-1999ms: 00.0000%
             50-179ms: 00.0000%   400-999ms: 00.0000%      2000+ms: 00.0000%
 <<<  Overall score:  9.4 / 10.0  >>>
Stopping TURN clients

[vi]

Tried running turnhammer though netem to reproduce faulty calculation of the badloss, but so far it is working properly for me.