I want to evaluate the maximum performance of a physical machine in a Redis scenario using the benchmark redisbench. Since I want to evaluate the overall throughput performance of the physical machine, I will start multiple Redis server processes (assuming 64 physical cores, I may start 64 or 128 instances) on the machine and simultaneously apply load testing to the overall throughput.

This will result in two problems. If I place the redisbench client process on the tested physical machine, it may compete for the computing resources of the server process, resulting in an inaccurate performance evaluation. If I place redisbench on another machine, it may be affected by network environments, which may result in inaccurate comparisons of test results between different machines (I only care about the CPU/memory capabilities of the tested single physical machine). A

re there any other better deployment methods to run the benchmark?