In the cloud storage, load balancing is a key issue. It would consume a lot of cost to maintain load information, since the system is too huge to timely disperse load. This paper presents a hybrid control strategy for load balancing. On the one hand the storage node cluster redistributes the load in its local range. On the other hand the system applies the overlapping structure to distribute the load to the global storage nodes by batch iteration approach. We also present a dynamic migrating strategy caused by two reasons: overload access or long response time in some storage nodes. Simulation results show that our approach can dramatically reduce the overall response time and better balance the global load.