Research And Implementation Of Conainer-based Cloud Resource Elasticity Strategy

In recent years,the container technology represented by Docker has become increasingly mature,and the function of its orchestration system Kubernetes has become more and more perfect.However,the existing elastic strategy of Kubernetes has some shortcomings.The existing elastic strategy of Kubernetes is analyzed,and the optimization strategies are proposed for the existing problems.The main work of this thesis is as follows:(1)When scheduling new Pods(the smallest resource scheduling unit in Kubernetes)based on the existing expansion strategy of Kubernetes,the problem of cluster load balancing and resource consumption costs is not considered after multiple Pods deployment.An improvement based on the resource scheduling strategy of ant colony algorithm is proposed to optimize the existing expansion strategy.Firstly,the objective function based on cluster load balancing degree and resource consumption cost is established.Secondly,the initial value of pheromone concentration and the termination condition of the classical ant colony algorithm are improved,and the problems of slow convergence speed and easy to fall into local optimum in resource scheduling are solved.Thirdly,the node selection process of the existing expansion strategy is optimized.Finally,the optimized elastic expansion strategy is implemented on Cloud Sim platform,and it is evaluated by comparing its performance against the basic scheduling algorithm(Bind)strategy,the classical ant colony algorithm(ACO)based strategy and the genetic algorithm(GA)based strategy.(2)In view of the existing shrinking strategy of Kubernetes,only the status priority of the Pod is considered when selecting the Pod to be deleted and the problem of node resource balance and node resource consumption costis not considered after the Pod is deleted.A shrinking strategy based on multi-dimensional resource balance and node resource consumption cost is proposed.The proposed two strategies are integrated and optimized with the existing one,so that the state priority of pod and the resource balance and resource consumption cost of the node can be considered comprehensively.Then,the optimized elastic shrinking strategy is implemented on the Cloud Sim platform,and compared it with the existing shrinkage strategy of Kubernetes through experiments.The results show that compared with the strategy of Bind,ACO and GA,the optimized strategy can reduce the cost of resource consumption and make the cluster load more balanced.Compared with the existing shrinking strategy,the proposed optimized elastic shrinking strategy can reduce the cost of node resource consumption,reduce the generation of node resource fragments,and make the use of node resources more balanced.