Research On Dynamic Migration Technology Of Container In Cloud Environment

With the rapid development of cloud computing and the wide application of container deployment,more and more attention has been paid to the research of container dynamic migration.Based on the short life cycle of container and high flexibility,there are many differences between container migration and traditional virtual machine migration.In this thesis,the corresponding algorithms of hot container selection and migration process are proposed,and algorithms are validated experimentally.This thesis mainly includes the following three aspects:(1)An optimized Minimal Migration Algorithm(OMNM)is proposed for the case where the actual container migration operation does not pay attention to the change of the container itself,which indirectly increases the number of migration after the container is moved out.The main idea of the algorithm is to start with the growth rate of the data volume of the container,to form historical data according to the size of each container at each time point for a period of time,and to use the fitting to judge the growth rate.If the growth rate is fast,priority is given to migrate container selection,while slower or slower data growth reduces prioritization and forms a mechanism for migrating alternatives.In the end,under the premise of ensuring cluster load balancing,resource waste is reduced.(2)The traditional pre-copy mode ignores the characteristics of the memory page,resulting in a situation in which frequently-modified memory pages are copied multiple times in the pre-copy phase.An optimized pre-copy algorithm(OPCA)is proposed to introduce the predictive model.According to the fact that some of the memory pages in the container are often changed,it is recommended to put these memory pages in the last copy of the downtime to copy those changes less memory pages.Finally,OPCA can reduce the number of iterations and downtime,improve the utilization of resources and enhance the user experience under the premise of guaranteeing that the total migration time is not higher than the traditional algorithm.(3)Design and implement a dynamic container migration system,in order to have better attention and development of containers in the dynamic migration.The core functions of the system include migration container selection module and container migration module which can meet the actual load balancing,user experience and other needs.Experiments show that the system can effectively complete the container migration in a short period of time and realize the load balancing of the cluster.To summarize,this thesis proposes optimized minimum migration algorithm and optimized pre-copy algorithm for hotpot container selection and migration process respectively.Through the simulation experiment and the realization of the prototype system,the proposed algorithm can effectively improve the efficiency of container migration and improve the utilization of system resources under certain conditions.