Dear @junjuew and @teiszler,
sorry for my late reply, but I had to put a little time out due to my health.
So, during that time and under the same network configuration, I launched Faceswap Server with Openstack ++ on a 6 CPUs and 4GB RAM, and the swap occurs with Latency of 50,3 ms and FPS of 18,42.
In that time, I have launch AWS Faceswap Server on Ireland with m4.large (2 CPUs and 8 GB RAM) and the test result is a latency 270-360ms and FPS of nearly 3,64.
What is your opinion and this result test?
Do you think better to test AWS Faceswap Server on Ireland with m4.xlarge (4 CPUs and 16 GB RAM)?
I do not understand clearly how I can manage and present the network transmission time and server computation time.
Thanks for your understanding and estimated help guys.