Massive concurrent access to servers is very common in cloud data centers. It is difficult for traditional network architecture to control traffic forwarding globally, and expensive load balancers are needed to deal with this application scenario. The SDN(software defined network) can globally control the network status through the controller, and use the switch as a load balancer, thereby reducing deployment costs. A server load balancing strategy based on OpenFlow switches was proposed in this paper. Service requests were partitioned and mapped through multi-address directed flow table, and the number of active connections was used as a load evaluation parameter. The optimal load redirection scheme was found through ant colony algorithm. During load migration, single-address directed flow table was used to ensure orderly forwarding of traffic at different stages. Experimental results show that the proposed strategy can effectively control the size of flow table and has better performance than traditional equalization strategies.