In data center network virtualization, the high coupling of the internal structure of the application often confines the considerations of most algorithms to node-to-node mapping, still unable to address isolation issues between physical nodes. The limitations of centralized mapping are apparent in cloud-native applications, as it fails to account for critical differences in virtual node requirements which is crucial for containerized microservices. This makes traditional algorithms ill-suited for practical network virtualization environments. To address this challenge, this paper proposes a distributed mapping algorithm, DCVNE, based on microservice applications. It allows a virtual node to be divided into multiple sub-nodes, with each sub-node responsible for only specific resource requirement and accepted by multiple physical nodes. Thus, it achieves joint utilization of physical node resources and more efficient load balancing. Compared with the most advanced solutions, the rich simulation experimental results indicate that DCVNE increases the acceptance rate of virtual requests by more than 10% while paying less cost.