The invention discloses a routing method for a data center network based on deep reinforcement learning. The method comprises the following steps of S1, carrying out regional division on the network,dividing the whole network into a plurality of communities, taking one community as one region, deploying an SDN controller on a node with the largest node betweenness in each region, and deploying anagent on the SDN controller; S2, establishing a global view of the network, establishing an adjacent matrix of the nodes by the SDN controller, and taking a resource value of a link between the nodesas a composite resource of the cache and the bandwidth; S3, constructing a route-oriented deep reinforcement learning framework; and S4, determining a path of the route and a forwarding rule, and making the SDN controller determine the route according to an action selection result of the agent. On the basis of the idea of data driving, the requirements of flow for the performance are converted into the requirements of flow for the network resources; and the multi-dimensional characteristic representation of the resources is found through deep learning perception and the capability of expressing high-dimensional data, the network resources are optimally distributed for the flow on an SDN control plane, and the self-adaptive intelligent routing is achieved.