The negative externalities from an individual bank failure to the whole system can be huge. One of the key purposes of bank regulation is to internalize the social costs of potential bank failures via capital charges. This study proposes a method to evaluate and allocate the systemic risk to different countries using a SIR type of epidemic spreading model and the Shapley value in game theory. The paper also explores features of a constructed bank network using real globe-wide banking data. The major findings are that the magnitude of the systemic risk at the national level is related to the degree distribution of a bank in a nonlinear fashion. To be more specific, it depends on whether the network is more heterogeneous such as a scale free network, or more homogeneous such as an exponential or even a regular network. The constructed global banking network includes over 30,000 public and private overseas banks all over the world. The systemic important institutions are identified.The detected modularity of the global network indicates that the geographical location still plays roles in formulating the communities. The systemic risk is internalized by capital charges required from each country. The capital charge is evaluated based on the country level systemic risk. A type-two-holling function is used to convert systemic risk to capital charge. Finally we suggest that individual risk control policy should be combined to the systemic risk control policy to maintain the stability of the system, neither of which can be ignored. This is an advice that is different from the current policy stance that emphasizes only the safety of individual bank.