Technical surfaces are composed of numbers of protrusions and grooves microscopically, however the appearances of the surfaces seem smooth. Surface roughness may be characterized by their intervals, periodicity, gradients and so on. We measured topographs of Fe-42 Ni alloys by atomic force microscopy. Surface roughness was evaluated in terms of power spectral density (PSD) which provides information on periodicity, and of slope histogram. It is found that a gradient of PSD is proportional to 1/f2, and the slope histogram is well fitted with the Gaussian distribution. The data of PSD show that the surface is naturally random with a strong correlation between neighborhood. It is found that our findings are explained by a simulated surface based on a random walk.