I try to measure the performance of different publish/subscribe systems, (rabbitMQ, apache kafka, pushpin,faye). In the first cycle, I try to measure the performance for 1000 requests with concurrency = 200. In the second cycle, I measure the performance for 10000 requests with concurrency = 200. However, I see that the throughput for the 2 measures do not change but it is stable in every pub/sub system.
For example, the throughput at rabbitMQ is:
- 1000-requests,200-concurrency => throughput = 256.33
- 10000-requests,200-concurrency => throughput = 261.24
Can you explain me why this is happens?