Based on the method of numerical simulation，this paper studies the impact of small worldliness on innovation performance of networks with low，medium and high openness in the dynamic evolution process of cluster formation and disruption． As many existing studies found the inverted U relationship between small worldliness and innovation performance，the study further finds that this inverted U relationship only appears in certain phases of network evolution． In networks with low openness，the inverted U relationship is significant in the phases where new clusters are being established，and not significant in the phases where former clusters are disrupting． In networks with high openness，the inverted U relationship is totally contrary． In the other phases of network evolution，small worldliness does not impact innovation performance significantly． This is because when innovation performance increases at a relatively high rate，individuals are more motivated to learn new knowledge． In networks with low openness，the increasing intervals are mainly in the phase where new clusters are being established，while in highly opened networks it is mainly in the phase where former clusters are disrupting． Regression result with the patent co-authorship data in China provides evidence for our conclusions． Policy implications of this study are that we should always keep a moderately clustered network，so that less redundant information are transmitted; Different levels of openness in different phases should be adopted，so that the advantage of different openness in different phases are taken． Organizations should actively react to emerging technology and make the adjustment of network relationship．