Network Load Resource Prediction Based On Hybrid Model Under Cloud Platform

In recent years,cloud computing has become the next major revolution in computer networks and Web supply.The combination of cloud computing and the Internet is also becoming increasingly close.The next-generation Internet represented by IPv6 will become one of the best options for future cloud computing.From the perspective of the cloud platform,as the Internet transitions to IPv6,more and more network users will emerge from IPv6,including direct users of the cloud platform and end users who provide application services on the cloud platform.From an internal point of view,a centralized and flexible service model for the cloud platform requires a large number of IP addresses,but it is limited by the number of IP addresses and address translation between internal and external networks,the deployment of the virtual environment and the entire infrastructure,and the network throughput capacity has been affected.The degree of limitation,thus affecting the performance of the entire cloud computing platform.Currently,cloud computing users enjoy the tremendous convenience brought by cloud computing,and at the same time,they also face some problems.The most important issue is the issue of resource allocation.The problem of resource allocation is mainly manifested in the following aspects:The traditional cloud computing model of fixed resource allocation based on configuration payment can no longer meet the growing needs of end users.Coarse-grained resource allocation methods are typical methods used by most cloud service providers today,such as Google and Amazon.This method takes the virtual machine as the scheduling unit,and according to the change of the load of the virtual machine,increase or decrease the number of virtual machines as needed to achieve the load adaptive function.However,this method of resource allocation often leads to problems such as distributed and scattered virtual machines,excessive occupation of physical machine resources,and serious waste of resources.With the increase in the number of cloud users,especially the increase in the number of mobile users,resource allocation methods are not only not conducive to improving resource utilization,but also lead to a substantial increase in operating costs.Therefore,how to improve the load balancing technology so that it can fully utilize the overall characteristics of the cloud platform,effectively manage resources,and maximize the use of resources is an issue that needs urgent solution and has practical application value.On the basis of in-depth analysis of cloud computing theory and load balancing technology,combined with the user characteristics of the cloud platform,this academic paper applies to the application scenario where the user request load on the cloud platform has a regular change,according to the different types of task load information,the required Resources are forecasted to provide guidelines for elastic scaling and on-demand scheduling of cloud platform resources.The main research work of this paper is as follows:(1)For the current methods of different load resource prediction,the ARIMA,Kalman,Support Vector Regression(SVR),Artificial Neural Network(BPNN)and Deep Learning(Deep)Learning)was summarized and summarized.(2)Considering different types of workload submitted by cloud users,this paper proposes a hybrid model-based forecasting method.The hybrid model combines Kalman filtering and an autoregressive integrated moving average model to predict the resources required for the workload.The experimental results show that this method has higher prediction accuracy than the single model prediction method,can effectively improve resource utilization,and is conducive to the on-demand planning of virtual machine resources.(3)Design and implement prototypes of resource prediction modules on open source cloud platforms to promote practical innovation with theories.The prototype of the prediction module was designed in the CloudSim simulation cloud platform,and a new real-time migration strategy was proposed to verify its effectiveness through experiments.