Complex network is adopted to describe the Chinese railway system, and two different ways are used to construct the networks. One of ways is that the stations are viewed as “nodes” and rails are regarded as “links”, thus a railway geographic network is generated. The geographic network has approximately zero average clustering coefficient, thus this network is a tree-form network. The other way is that by viewing stations as “nodes”, an arbitrary pair of stations is considered to be connected by a “link” when at least one train stops at both stations. Thereby, a railway traffic network is generated. The traffic network has comparatively large average clustering coefficient (〈C〉=0.83) and small average path length(〈d〉=3.27), and the degrees of most nodes of the traffic one follow a scale-free degree distribution. Consequently, the traffic network is a small world network with scale-free property.