There is of course the official documentation: https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem
However, my rule of thumb is to go 10x faster than the expected signal rate when acquiring data. With this you could still distinguish 9.0 and 9.1 GHz.
But 9 GHz ... that's really fast signal. I don't think there are any signal acquisition boards from NI available for that. For that speed I guess you need a device between the acquisition and the sensor. I tempted to advice you on radar acquisition devices. Those aren't cheap, but do the trick. Call someone at Keysight. I guess they can do also FFT and spectrum analysis tools, saving you time on the data analysis.
