| Network Function Virtualization(NFV)technology decouples network functions from proprietary hardware devices.By using virtualization technology,Virtual Network Function(VNF)runs on x86 general-purpose servers and orchestrates services through service orchestration.The software manages VNFs and orchestrates services,which improves the flexibility of the network and reduces the capital cost and operation and maintenance cost of telecom network providers.In order to improve the utilization rate of network function infrastructure,it is necessary to allocate resources flexibly to network function instances according to service traffic.The development of virtualization technology and disk I/O technology has greatly enhanced the performance of VNF.According to the type of virtualization,VNF has two different carriers:virtual machine and container.These two virtualization technologies are used in resource isolation.Due to the short container startup time and resource restriction mechanism,the original NFV based on virtual machine scaling policy does not apply to the container-based NFV platform.The expansion and contraction include four steps:monitoring,analysis,planning and execution.This paper studies the latter three steps,and the specific research results are as follows.1)Based on the LSTM periodic traffic forecasting capacity reduction and expansion mechanism,the capacity reduction and expansion strategy is formulated according to the data monitored by the traffic forecast analysis.Aiming at the problem that there are two modes of periodic traffic and burst traffic in network traffic,which are difficult to predict,this paper designs a capacity reduction and expansion mechanism based on LSTM periodic traffic prediction.The advantage of time series prediction is to use the periodic extraction algorithm to process historical data,use the preprocessed data to model the business traffic,and then use the prediction model to predict the traffic and make a judgment of scaling and expansion according to the predicted value.Simulation experiments show that the capacity reduction and expansion mechanism based on LSTM periodic traffic prediction can improve the accuracy of traffic prediction on the one hand,and avoid the impact of burst traffic on the capacity reduction and expansion strategy on the other hand.2)Based on container-based VNF implementation and elastic resource allocation,study the strategy of scaling and expanding capacity using container virtualization technology.The container implements virtualization at the level of the host operating system.It has the characteristics of rapid startup,small image size,and the ability to adjust resource usage at runtime.In order to study the impact of these new features on scaling,this paper implements a module based on the Click project.The VNF container instance is created,a container management platform is established based on the Docker open source project,and the API provided by Docker is encapsulated to realize the management of the full life cycle of the VNF in the form of a container.The elastic scaling scheme extracts traffic characteristics and uses different scaling and scaling strategies for traffic in different modes.Simulation experiments prove that horizontal expansion and vertical expansion can flexibly cope with the coexistence of burst traffic and periodic traffic,and ensure service SLA. |
