[srslte-users] Inverse sawtooth wave behaviour of RTT latency

Fri May 29 17:25:47 UTC 2020

Dear srsLTE team and users,

I’ve come to notice a strange behaviour of UE latency measurements of the srsLTE stack using LaTe, a ping-like tool.
Basically, what happens is that RTT latency shows a inverse sawtooth wave behaviour, that changes consistently based on the probing frequency.

To give a quick example, if the probing frequency is an integer multiple of 100 ms, e.g. 200 ms, then the latency of a given test is going to be more of less constant. If instead a value of frequency close to a multiple is chosen, let’s say 201 ms, then I’m observing a slowly decreasing pattern, where each probe measures approximately 1 ms lower latency that the previous one. After the latency goes to a minimum of about 18 ms, it goes up againt to about 36 ms, as an inverse sawtooth wave.  If I choose 202 ms, same behaviour, but now the difference is 2 ms (as you can read in the report below).

What could be the cause of this behaviour?
My idea is that it could be something periodic, maybe scheduling related. Can it be the LTE MAC scheduler? It feels strange since it operates at TTI scale, unless we consider the possibility of aliasing. But then anything could be (one of) the cause(s): Linux thread scheduler, or even USB polling frequency.
If you have any idea I’d be more than happy to hear it.

Below you can find the setup description and a latency report where you can observe the behaviour I explained before.
I’m using 2 USRP B210, connected to each other with cables and attenuators (so there should be no external interference).
The boards are connected to Intel i7-7700HQ laptops, which run Ubuntu 18.04. SrsLTE 19.09 is tested, with UHD 3.15. 
The test is performed using LaTe, a latency tester tool that works very similarly to ping, but with some additional features (https://github.com/francescoraves483/LaMP_LaTe).

[INFO] New periodic interval: 200.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=20). Time: 27.193 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=21). Time: 27.207 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=22). Time: 27.223 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=23). Time: 27.155 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=24). Time: 27.182 ms (User-to-user)
[INFO] New periodic interval: 202.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=25). Time: 27.410 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=26). Time: 25.154 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=27). Time: 23.201 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=28). Time: 21.213 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=29). Time: 19.179 ms (User-to-user)
[INFO] New periodic interval: 201.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=30). Time: 37.483 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=31). Time: 36.107 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=32). Time: 35.139 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=33). Time: 34.156 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=34). Time: 33.078 ms (User-to-user)
[INFO] New periodic interval: 202.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=35). Time: 32.191 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=36). Time: 30.027 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=37). Time: 28.118 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=38). Time: 26.086 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=39). Time: 24.130 ms (User-to-user)
[INFO] New periodic interval: 201.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=40). Time: 21.985 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=41). Time: 21.015 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=42). Time: 20.027 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=43). Time: 19.068 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=44). Time: 18.067 ms (User-to-user)
[INFO] New periodic interval: 200.000 ms
Received a reply from 172.16.0.1 (id=57906, seq=45). Time: 37.107 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=46). Time: 37.074 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=47). Time: 36.987 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=48). Time: 37.017 ms (User-to-user)
Received a reply from 172.16.0.1 (id=57906, seq=49). Time: 37.000 ms (User-to-user)

Thank you
Best regards
Corrado Puligheddu

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.srsran.com/pipermail/srsran-users/attachments/20200529/f0ae7e80/attachment.htm>