Is it possible to limit the amount of CPU resources a C API is allocated when running? We have an API that when running on a test server grabs over 40% of the CPU resources and we don’t have that amount space on the production environment so are unable to test the true performance of the API.