P.S
_The error percentage is not 0 in the second image. It's because of the ulimit. This benchmarking was done in Mac with ulimit 4096. This is not enough to handle all those concurrent request. But as you can see there was no error(TCP ) in the first image with 100 concurrent clients doing 10K requests. So it meanse if we run this test in a Linux server with ulimit 65,536 (this is the default one) we more likely won't observe any _