Dynamic Migration Method of Virtual Network Functions Based on Deep Belief Network Resource Demand Prediction

Abstract

The invention relates to a virtual network function dynamic migration method based on deep belief network resource demand prediction, which belongs to the field of mobile communication, and includes step S1: aiming at the dynamic characteristics of SFC service resource demand in a slice network, establishing a comprehensive migration overhead and bandwidth overhead System overhead model; S2: In order to realize spontaneous VNF migration, monitor the resource usage of virtual network functions or links in real time, and use the method of adaptive DBN prediction based on online learning to timely discover the underlying nodes or links deployed in it. Resource hotspots; S3: Design a topology-aware dynamic migration method based on the prediction results to reduce system overhead; S4: Propose an optimization method based on tabu search to further optimize the migration strategy. The prediction method of the invention not only speeds up the convergence speed of the training network, but also has a good prediction effect. Combined with the migration method, it effectively reduces the system overhead and service level agreement violation times, and improves the performance of network services.

Description

technical field [0001] The invention belongs to the technical field of mobile communication, and relates to a virtual network function dynamic migration method based on deep belief network resource demand prediction. Background technique [0002] At present, the mobile network industry is rapidly evolving to 5G, and the three new application fields of "mobile broadband enhancement", "large-scale Internet of Things", and "low latency and high reliability communication" will play an important role. The 5G network has high flexibility to cope with the business changes of mobile operators, especially the concept of network function virtualization enables the infrastructure to flexibly meet the diversification of vertical application requirements. Network slicing is a technology for flexible configuration of resources in wireless virtual networks, which can be quickly deployed and managed centrally. It mainly uses software defined network (Software Defined Network, SDN) and netw...

AI Technical Summary

Problems solved by technology

Existing inventions have proved that neural network technology can well predict the relationship between resource characteristics and resource requirements. Although it shows that the prediction accuracy of neural network is higher than that of traditional statistical models, it does not involve the existence of neural network in the prediction process. Problems such as long training period, slow convergence speed and easy to fall into local minima

Images