The aquaglyceroporin GlpF is a member of the aquaporin family. It selectively conducts small molecules, such as glycerol, across the cell membrane under a concentration gradient of the substrate. Atomistic molecular dynamics (MD) simulation would provide great insight into the substrate transport mechanism of GlpF and membrane channels alike. Ideally, non-equilibrium simulations under various concentration gradients of glycerol are desired to emulate the transportation in cells, but this kind of simulation is difficult due to a complicated system setup and high computational cost. Here, we present a new strategy to extract non-equilibrium kinetic information from equilibrium MD simulation. We first performed long-time (totally 22.5 μs) multi-copy equilibrium MD simulations of glycerol conduction through GlpF. Tens of times the spontaneous permeation of glycerol through GlpF was observed, allowing us to elucidate the detailed mechanism of the stereoselectivity for glycerol. Then we employed Markov state model (MSM) analysis of the MD trajectories to identify the intermediate states during glycerol transport and calculate the inter-state transition rate constants. Based on the results of MSM analysis, we built the kinetic models of glycerol transport and calculated the glycerol fluxes under various concentration gradients by solving the master equations. The results agree well with the experimental measurement at a certain glycerol concentration, and provide holistic information on the glycerol conduction capacity of GlpF. Our work demonstrates that long-time atomistic MD simulations can now bridge the microscopic dynamics and the kinetic description of substance transport through membrane channels, hopefully facilitating the engineering of new selective channels for various molecules.