RTT time is incorrect while using multithreading
s1042992 opened this issue · 1 comments
s1042992 commented
In python, I created a thread list and append the 40 times ping works measure_latency(host, port) into the list, then the result was like:
CH. 3 = 15.075 ms
CH. 17 = 20.111 ms
CH. 16 = 20.111 ms
CH. 4 = 20.593 ms
CH. 9 = 20.111 ms
CH. 14 = 21.094 ms
CH. 6 = 20.595 ms
CH. 8 = 20.595 ms
CH. 7 = 24.604 ms
CH. 11 = 20.111 ms
CH. 13 = 24.606 ms
CH. 10 = 25.609 ms
CH. 12 = 20.094 ms
CH. 2 = 25.104 ms
CH. 15 = 25.107 ms
CH. 35 = 25.105 ms
CH. 32 = 29.118 ms
CH. 1 = 25.104 ms
CH. 29 = 29.117 ms
CH. 24 = 33.128 ms
CH. 5 = 35.132 ms
CH. 31 = 29.117 ms
CH. 20 = 34.632 ms
CH. 21 = 35.633 ms
CH. 22 = 33.628 ms
CH. 36 = 34.129 ms
CH. 30 = 38.139 ms
CH. 26 = 38.641 ms
CH. 18 = 38.641 ms
CH. 25 = 39.142 ms
CH. 33 = 36.134 ms
CH. 28 = 38.641 ms
CH. 40 = 43.153 ms
CH. 34 = 35.132 ms
CH. 37 = 44.158 ms
CH. 23 = 25.607 ms
CH. 38 = 42.652 ms
CH. 39 = 44.158 ms
CH. 19 = 34.632 ms
CH. 27 = 39.142 ms
The RTT time above somehow progressively increases.
But without multithreaded, the result will be like:
CH. 1 = 22.148 ms
CH. 2 = 10.528 ms
CH. 3 = 11.267 ms
CH. 4 = 20.77 ms
CH. 5 = 22.03 ms
CH. 6 = 20.608 ms
CH. 7 = 22.036 ms
CH. 8 = 21.07 ms
CH. 9 = 11.059 ms
CH. 10 = 10.388 ms
CH. 11 = 11.491 ms
CH. 12 = 21.23 ms
CH. 13 = 21.141 ms
CH. 14 = 10.883 ms
CH. 15 = 21.39 ms
CH. 16 = 10.012 ms
CH. 17 = 11.529 ms
CH. 18 = 22.938 ms
CH. 19 = 21.064 ms
CH. 20 = 10.528 ms
CH. 21 = 20.711 ms
CH. 22 = 21.616 ms
CH. 23 = 21.658 ms
CH. 24 = 20.516 ms
CH. 25 = 21.279 ms
CH. 26 = 10.48 ms
CH. 27 = 22.067 ms
CH. 28 = 21.048 ms
CH. 29 = 10.035 ms
CH. 30 = 21.028 ms
CH. 31 = 11.028 ms
CH. 32 = 21.891 ms
CH. 33 = 21.369 ms
CH. 34 = 10.487 ms
CH. 35 = 21.774 ms
CH. 36 = 20.57 ms
CH. 37 = 11.006 ms
CH. 38 = 10.755 ms
CH. 39 = 21.569 ms
CH. 40 = 10.053 ms
It seems very normal without multithreaded.
I don't know why the result is incorrect in multithread, any help would be appreciated!
ATolkachev commented
@s1042992 it could be caused by GIL. Better to use multiprocessing instead of multithreading.