Network Function Virtualization (NFV) enables flexible deployment of network services as applications. However, it is a big challenge to guarantee the Quality of Service (QoS) under unpredictable network traffic while minimizing the processing resources. One typical solution is to realize NF scale-out, scale-in and load balancing by elastically migrating the related traffic flows. However, it is difficult to optimally migrate flows considering the resources and QoS constraints. In this paper, we propose DeepMigration to efficiently and dynamically migrate traffic flows among different NF instances. DeepMigration is a Deep Reinforcement Learning (DRL)-based solution coupled with Graph Neural Network (GNN). By taking advantages of the graph-based relationship deduction ability from our customized GNN and the self-evolution ability from the experience training of DRL, DeepMigration can accurately model the cost (e.g., migration latency) and the benefit (e.g., reducing the number of NF instances) of flow migration among different NF instances and employ dynamic and effective flow migration policies generated by the neural networks to improve the QoS. Experiment results show that DeepMigration reduces the migration latency and saves up to 71.6% of the computation time than the state-of-the-art.

网络功能虚拟化（NFV）可以灵活地将网络服务作为应用程序进行部署。然而，在最小化处理资源的同时，在不可预测的网络流量下保证服务质量（QoS）是一个巨大的挑战。一种典型的解决方案是通过弹性迁移相关业务流来实现NF横向扩展，横向扩展和负载平衡。但是，考虑到资源和QoS约束，很难最佳地迁移流。在本文中，我们提出了DeepMigration以在不同NF实例之间高效，动态地迁移流量。DeepMigration是基于深度强化学习（DRL）的解决方案，并与图形神经网络（GNN）结合使用。通过利用我们自定义的GNN中基于图的关系推论能力和DRL经验训练中的自演化能力，DeepMigration可以准确地建模成本（例如，迁移延迟）和收益（例如，减少数量NF实例）在不同NF实例之间进行流迁移，并采用由神经网络生成的动态有效流迁移策略来提高QoS。实验结果表明，与现有技术相比，DeepMigration减少了迁移延迟，并节省了多达71.6％的计算时间。