The invention relates to a data recovery method for distribution optimization of data storing nodes in cloud storage. A storage strategy enabling locally optimal distribution of data and data copies in data nodes is provided; a decision domain is established by using a value that a data object is mapped into an annular Hash value space as a computing center; the weights of a load rate, a relative load temperature and predicted waiting time serve as decision criteria to guide establishment of decision distribution, and an optimal storage node which is selected; and establishment of an effective decision set, expansion of the decision set and addition and deletion of nodes after data change are included. According to the data recovery method, the probability of non-uniform node distribution is reduced.