Finding the point where the SUT starts to level out on response per second ans strange load model behaviour

I am running this load model:


setUp(scn.inject(rampUsersPerSec(0.1) to (30) during (5 minutes)).protocols(httpConf))


and I increase the value of 30 above to try to find the point where my SUT starts to level out on responses per second/where it can’t handle more load without an increase in response time.
Question 1: is this a right approach to finding the number of users my SUT can handle before response times starts to decrease?
Question 2:
The number of active users remains very low for almost the entire scenario (see attachment number of active users) but spikes at the end. Why is that. I thought my load model would add more-and-more users along the simulation when I use this load model.