In this talk, we summarize our recent efforts in understanding the global function and dynamics of cellular networks. We found that the dynamics of the cellular networks can be determined by the two driving forces. One is the gradient of the underlying landscape and the other is from the curl flux. The underlying landscape is linked to the probability distribution of the steady state and provides a global picture for describing the cellular networks. We found that the landscape can be used to quantify the global stability and robustness of the system. The non-zero flux breaks the detailed balance and therefore gives a quantitative measure of how far away the system is from the equilibrium state, reflecting the degree of the energy input to the system. Our decomposition of the driving forces of the complex systems into landscape gradient and curl flux establishes the link between the dynamics and the underlying thermodynamic non-equilibrium natures. We applied our theory to several cellular networks such as budding yeast cell cycle, cell fate decision making in differentiation and reprograming. For yeast cell cycle oscillations, we found the underlying landscape has a Mexican hat ring shape topology. The height of the Mexican hat determines the global stability. The landscape gradient attracts the system down to the oscillation ring. The curl flux is the driving force for coherent oscillation on the ring. The landscape topography is crucial for coherence of the oscillation and can be used to identify the hot spots in the underlying cellular networks for function of cell cycle. We are able to quantitatively map out the landscape of differentiation and development. We are also able to quantify the speed and paths for differentiation and reprograming, important for regeneration medicine and tissue engineering.