Research On Elastic Services Of Edge Layer On Edge Computing Platform Based On Kubernetes

With the continuous increase of mobile services,the traditional architecture has been unable to meet the increasing demand of mobile services.Cloud computing has become the focus of the core network architecture.However,with the advent of the 5G era,ultra-high bandwidth,ultra-high density and ultra-low latency will become the main characteristics of services.The speed,delay and stability cannot be guaranteed at all if cloud computing is still the only computing mode.Edge computing is a computing mode that provides services to users at the edge of the network,which has the characteristics of low latency and high bandwidth.This mode can make users enjoy a stable and high-quality network experience.At the same time,with the emergence of virtualization technology,the traditional server resources can be more fully utilized.Platform and application virtualization is rapidly moving to the edge of the cloud,paving the way for the multi-core processors in the core of the Internet of things,mobile and embedded application devices.In this thesis,an edge computing platform is built by integrating container technology called Docker and containers management technology called Kubernetes.In order to make up for the shortcomings of elastic algorithm on edge layer,the original elastic algorithm is improved and a CPU-Memory Crossed Maximum algorithm is proposed in this thesis.The algorithm takes into account the requirements of the application for memory and CPU,and makes a decision on the number of target Pods according to memory and CPU.The algorithm is tested in the Kubernetes platform,and compared with the original algorithm.The results show that the CPU-Memory Crossed Maximum algorithm can provide a more accurate number of target Pods while ensuring the stable scaling work.On the basis of the improved elastic algorithm,the network service experiments on the edge computing platform are carried out,and the stability and delay advantages of the edge layer to the cloud layer are verified by comparing the network indicators on both layers.In addition,the design and construction process of edge computing platform are described in detail in this thesis.In order to make full use of the resources on the cloud layer and the edge layer,a machine learning idea of training on cloud layer and predicting on edge layer is also proposed in this thesis.The feasibility of the idea is verified through the successful implementation of image recognition service on the edge layer.