The survivability problem of large scale time synchronization network is studied. With autonomous configuration of the nodes, the network topology is not static and hierarchical but dynamic and self-organizing. A new concept of search world is presented, and the BA model of time synchronization network based on it is built. The model is also optimized to make the network more survivable. After failure of some nodes, other effected nodes can be self-healing. The self-healing algorithm is given to mitigate the load of the network caused by self-healing nodes.