> The test is structured as a continuous 1-hour ramp up from 0 to 300k devices.<p>Number of devices would make a far clearer graph axis than minutes. It's not obvious what you're doing when one axis is milliseconds and the other is minutes.