To avoid monster community problem which suffered by distributed k-clique community detection, τ-window community detection was proposed. In addition, τ-window centrality estimation was put forward. By investigating the periodic evolution of τ-window community and τ-window centrality, two new metrics, TTL (time to live) community and TTL centrality, were proposed to improve the prediction of the node's encounter during the message's lifetime. Moreover, a social-aware routing algorithm, PerEvo, was then designed based on them. Extensive trace-driven simulation results show that PerEvo achieves higher message delivery ratio than the existing social-based forwarding schemes, while keeping similar routing overhead. ©, 2015, Tongxin Xuebao/Journal on Communications. All right reserved.