Abstract: The number of replicas in the control strategy of a traditional distributed file system is calculated based on internal resources while changes in external demand are ignored. However, this strategy is not suitable for deployment in a service-based, resource-rich internal storage "smart city" cloud storage center. We propose a control model for the number of replicas ,which combines data security(the minimum amount of copies)together with service needs (best copy volume). A predictive algorithm based on double time granularity is included in this model to predict the popularity of a file. In addition, the number of copies in the cloud adjusts itself dynamically according to the popularity of a file and system resources. Simulation experiment results show that the accuracy of double time granularity forecasting is greatly increased. Compared to the original static copy, a dynamic copy mechanism based on double time granularity prediction has a significant advantage in response to sudden large-scale concurrent accesses and its storage resources occupancy rate is low.