Abstract: To solve the problem of the task scheduling result affected by the state change of server in the cloud computing system, a scheduling algorithm for the cloud computing system with the dynamic feedback of server state (DFSS) was proposed. Firstly, a sliding window model was proposed for the scheduling algorithm to reflect the server busy state in real time. And then, a weight curve model of time window and an average utilization rate model were used to compute the average utilization rate and improve the predict accuracy of the server. Finally, a server feedback weight strategy was proposed based on the average utilization rate to calculate dynamic rank value of every task node, and all task nodes were sorted by their dynamic rank value in order and respectively schedule them. The experiment result indicates the proposed algorithm can effectively reduce the task Makespan.