Running load testing on this setup is slower than running directly on node worker

[dariobanfi]

Hi,

First of all thanks for the example and the great article. I configured the project and tried to run some load testing tools (siege, autocannon) on my machine to see the difference, but I actually noticed that it's much slower with the nginx+haproxy+cluster setup than running directly on one single worker instance.

autocannon localhost:9000/health caps at around 200 req/s autocannon localhost:9002/health caps at around 5110 req/s (the worker process)

Did I misunderstand something or is it a problem of my computer?

Thanks