Research On Load Distribution Optimization For VoLTE Networks Based On Traffic Prediction

After the development of mobile voice from TDM to IP,from traditional switch to softswitch,it will move to mobile broadband voice based on IMS in the future.In terms of network evolution,the wireless side develops from GSM/CDMA/UTMS to LTE,while the core network side develops from CS to IMS.IMS has gradually become the core network standard architecture in the all IP era because it supports multiple access and rich multimedia services.At the same time,compared with CSFB and SVLTE relying on CS domain to provide voice,IMS can provide voice solution with QoS based on LTE,which has become the inevitable result of the development of wireless and core network technology.With the continuous increase of the number of IMS based VoLTE network users and the continuous enrichment of business application scenarios,IP data traffic is also showing a rapid growth trend.IP Bearer Network is the main carrier network of VoLTE services.With the rapid development of services,the problems of network congestion and low quality of service are becoming more and more serious.In order to provide better service quality for VoLTE users and reasonably allocate resources and loads,it is necessary to optimize the existing QoS deployment strategy of IP Bearer Network,and at the same time,it is necessary to accurately predict the network traffic and predict when the tide will occur.In this paper,we propose the problem of flow prediction around the tide and under the background of VoLTE service IP Bearer Network,we optimize the network configuration and predict the network traffic from the following ideas.(1)This paper studies the IP Bearer Network of VoLTE service based on IMS,discusses the problem of traffic tide and its causes,and tries to optimize the QoS of IP Bearer Network,so as to improve the support for optimizing load deployment through traffic prediction.In order to obtain the packaged traffic data for deep learning,we try to establish the IP bearing network model which is in line with the tide phenomenon.(2)This paper analyzes the problems and limitations of the previously commonly used linear regression method in flow prediction.Aiming at the plagiarism of traffic in metropolitan area network,firstly create an appropriate BP-ANN neural network and set appropriate relevant parameters;In addition,under the premise of attention mechanism,an appropriate LSTM network is designed to deal with the temporal correlation of traffic data.According to these two neural network models,the traffic of metropolitan area network is predicted,and the tidal migration period is well predicted.Find suitable iteration times and other parameters in the model to continuously reduce the error of flow prediction in the model,and finally make the traffic prediction accuracy strive to reach 85% or more.(3)The feasibility of this scheme in the actual network is preliminarily demonstrated by using the method of actual measurement,and the effectiveness of the scheme in the network access side is verified.