I have a simple grpc service that returns "hello" to the client. The server is running on 80 core machine. I am using ghz tool to load test the service. It's able to serve 40k requests per second but the rest of the 20k requests are queued and all grpc threads are in waiting state. CPU is underutilized to 7% and memory is 7-10%.
More details are mentioned here:
What are the required settings to support more load of incoming requests?