Cooperative training method for multiple SDN (Software Defined Network) controllers

Abstract

The invention discloses a multi-SDN (Software Defined Network) controller cooperative training method. The method comprises the following steps: step 1, sending a model optimization request; step 2, sending a federal learning notification; step 3, uploading information of the edge SDN controller; 4, distributing the initial model and standard parameters; 5, the edge SDN controller carries out model training; step 6, the edge SDN controller uploads the model; step 7, the central SDN controller dynamically adjusts the model aggregation weight, completes global model aggregation and calculates the accuracy level; and step 8, if the global model reaches the accuracy level requirement, returning an optimization result to the network composer, otherwise, starting the next round of federal learning training. According to the invention, through cooperative control of the edge nodes and the central controller, the accuracy of a central node network topology model is effectively improved, and then the reasonability of path selection and flow load mean values is improved.

Description

technical field [0001] The invention belongs to the technical field of mobile communication networks, and in particular relates to a multi-SDN controller cooperative training method. Background technique [0002] In the cloud-network integration scenario, the SDN master controller not only manages and controls each network node device, but also manages the virtual machine network in the cloud management platform. In the scenario where multiple SDN controllers perform federated learning, some node SDN controllers may fail to meet the accuracy requirements within the specified training time, or overfitting occurs due to the small local data set. The above situations will cause the global model aggregated by the central SDN controller to fail to achieve high accuracy, thereby affecting the overall performance of the system. [0003] Existing method 1: Traditional federated learning algorithms, such as FedAvg, FedProx, etc., the central SDN controller aggregates the models uplo...

AI Technical Summary

Problems solved by technology

[0005] However, the above method, the existing method 1 will undoubtedly greatly reduce the accuracy of the global model; the existing method 2 cannot be applied to the scene where the central SDN controller does not have global data

Images