Research And System Implementation Of Container Placement And Migration Methods In Cloud Computing Environment

With the rapid development of cloud computing in recent years,Containers as a Service(CaaS)has gradually become a Service mode,and the scale of container cloud keeps increasing,resulting in the problem of large resource consumption and low resource utilization of container cloud.As virtualization technology is the core technology of cloud computing,the research of container placement migration algorithm is of great significance to solve this problem.Therefore,this paper studies container placement and container migration respectively to solve the above problems.The main research contents are as follows:In order to solve the problem of large amount of computation and low efficiency in the process of container placement using heuristic algorithm,a clustering container placement algorithm based on weighted mean shift was proposed.Firstly,a communication model is established based on the saturation ratio of communication links and network topology.The communication cost between each physical machine is calculated by the model,and the physical machine with lower communication cost is selected as the physical machine to be allocated.Then,the weighted mean shift clustering algorithm proposed in this paper is used to improve The FFIGA algorithm.In the crossover mutation stage,the weighted mean shift clustering algorithm is used to cluster the physical machines,and only the physical machines belonging to different clusters are crossed and mutated,so as to improve the efficiency of crossover mutation and improve the computational efficiency of the algorithm.In the pre-copy container migration algorithm,the container downtime and total migration time are prolonged because the static stop condition is used in the cyclic copy memory phase and the read/write mode of memory is not considered.To solve the above problems,a container migration algorithm based on autoregressive comprehensive moving average was proposed.First of all,experiments are carried out on loads with different write memory rates to verify that the write dirty frequency of memory pages is different under different loads.Then,the autoregressive comprehensive moving average algorithm is used to predict the proportion of dirty pages in the container memory,determine the read/write mode of the container memory,and dynamically determine the stop condition of the cycle iteration according to the read/write mode of the container memory,which speeds up the container migration,shortenes the container downtime and the total migration time.The container placement and migration system under cloud computing is designed and implemented.The system is mainly composed of login and user management module,physical machine cluster management module,container management module,container placement module and container migration module.The system passed the functional test and the performance test met the requirements of requirement analysis.When the number of concurrent requests was more than 500,the average response time was less than 1400ms.