The invention provides a system volume analysis and prediction method and device. The method comprises the steps of obtaining system operation data, establishing a relationship between system condition data, business state data and abnormal fault data, and using the data as the foundation to establish a volume regression model, and on the basis of the volume regression model, calculating predictive volume use data of all hardware resources in a designated time and system performance data in the designated time, based on the two types of data, referring to the volume regression model, giving the prediction to the subsequent hardware resource volume, therefore through the comprehensive analysis of historic operation state data, analyzing the volume use conditions of all key resources of all stages, predetermining whether the volume of the hardware resources such as servers reaches a bottleneck, and giving warnings to potential volume risks to avoid the situation that the repair is only implemented after a hardware resource fault occurs. Therefore, the system safety is enhanced.