iperf3: The server grace period is too short and should be passed in as an argument

[HeFeng1947]

Bug Report

The server grace period is too short causing the socket to be closed in advance. Related submissions #859

Why I need long grace period

I use iperf3 to test the throughput test and network stress test of thread network (a low power 802.15.4 mesh network based on IP stack). Test steps can be refer to this issue openthread/openthread#7572

Due to the difference between the transmission rate of the UDP layer and that of the actual wireless physical layer(less than 7kB/s), the client ends sending quickly but it takes a long time for the server to receive all the UDP packets.

Currently, the grace period on the server side is hard-coded at just 40 seconds. This often causes the server to end reception prematurely. I think the grace period needs to be a flexible parameter.

[bmah888]

Thanks for your comments! I understand where a longer grace period is something that'd be useful for your tests. For the environments that iperf3 was designed for (high-speed R&E networks), 40 seconds is actually a very long time, and I'm not convinced that tuning this is generally needed for the use cases we prioritize. I'd recommend for your situation that since you have the source code, you can just make the change locally in your iperf3 sources (forking the code repo if necessary) and build a custom version.

[davidBar-On]

just adding a note: if it is known that the interface throughput is slow (7Kbps), then sending bit rate can be limited so not all the UDP packets will be sent at once. This should be done by both limiting the bit rate, e.g. -b 10K, and limiting the message length, e.g. -l 512 (as the UDP default message size is 32KB).