With the rise of Internet economy, network virtual community has become an important way of knowledge accumulation. However, it encounters various kinds of knowledge sharing predicaments at different development stages. First of all, based on two assumptions of “asymmetric information” and “collective rationality”, we derive the optimal knowledge output level of virtual community and establish the baseline of the research. Furthermore, network virtual community is divided into two development stages which are initial stage and mature stage. For governance needs, “screening mechanism” and “incentive mechanism” are adopted respectively to promote its knowledge sharing level. The study shows that the optimal reward for a “knowledge talent” should be positively related to its amount of knowledge sharing and unit cost, while at the mature stage this reward should also be related to the number of members; the optimal reward for a “knowledge rookie” at the initial stage should be the expected benefit gap between two types of members, while at the mature stage this reward can be zero.