In recent years, sampling inequalities have become increasingly important to prove convergence results for various numerical methods based upon meshfree discretisations. They are usually stated in a Sobolev space setting and provide results of the form that if such a function is known or controlable at a discrete point set and if a high-order norm is also controlable, then all norms up to that particular norm are controlable. In a certain way, they are generalisations of classical Poincare-Friedrichs inequalities. They are the quasi standard for providing error estimates for kernel-based approximation. In this talk I will discuss such sampling inequalites for variations of tensor product Sobolev spaces comprising, amongst others, mixed regularity Sobolev spaces and Griebel-Knapek spaces. They are important if functions are approximated, which are defined on a high-dimensional domain and/or have different smoothness properties for different coordinate directions, which is, for example, the case for the solution of time-dependent PDEs.